6G RF PCB Design: Sub-THz Constraints Reach Phone Programs Early

On September 4, 2026, industry reporting indicated that European CE-RED requirements are extending radio frequency compatibility validation for high end AI smartphones further into the 6G sub-terahertz bands, and are bringing RF front end modules, full-device coexistence testing and certification documentation into an earlier validation stage. The same reporting described pre-scan requirements covering 0.21 to 0.275 THz and out-of-band spurious emission testing across 100 to 300 GHz.

An important qualification applies here. No verifiable European Union official document has been located that matches a claim of a uniform mandatory Sub-6GHz plus sub-THz pre-scan requirement for all AI phones from a specific date. It is more accurate to read this as an industry signal that European RF compliance is extending toward higher bands, rather than a confirmed and finalized mandatory milestone. The EU has this year formally advanced spectrum coordination for the 116 to 260 GHz range, establishing a regulatory foundation for devices operating above 100 GHz.

Why PCB Design Has to Consider Sub-THz Before 6G Ships

Commercial 6G deployment is still years away, but RF hardware development cannot wait for standards to be frozen. A phone passes through multiple iterations between concept design, RF validation, certification and volume production. The antenna, power amplifier, filter, switch and PCB together determine the final RF performance, and changing any of them late in the process invalidates the validation work already completed.

Once spectrum extends above 100 GHz, manufacturing deviations that had limited effect at Sub-6GHz are amplified considerably. The reason is straightforward: the higher the frequency, the shorter the wavelength. Trace width, copper surface roughness, dielectric thickness and even the parasitic parameters introduced by pads and vias all have a greater effect on the signal.

Problems that could previously be corrected by post-production tuning on a phone mainboard may have to be resolved during layout, material selection and stackup definition instead. That is the practical meaning of the shift for PCB engineering. The design stage absorbs constraints that used to be handled during integration.

Materials and Impedance Consistency Come First

RF PCBs have long been specified around 50 ohm impedance, insertion loss and return loss. Those metrics remain important at higher frequencies, but the tolerance available for manufacturing error narrows significantly.

Consider dielectric properties. If the dielectric constant and loss tangent of a laminate vary substantially between production batches, the same design will exhibit different electrical lengths on different batches of board. Length differences that are irrelevant at lower frequencies translate into phase differences at sub-terahertz frequencies, and phase differences degrade antenna matching and beam steering accuracy.

Copper foil roughness behaves similarly. An increase in surface roughness raises conductor loss, and at high frequencies conductor loss is no longer negligible relative to dielectric loss. The two losses must be balanced through material and process selection together.

For a complex RF front end, trace length variation, reference ground discontinuity or differences in via structure can all affect phase and antenna matching. This is why the value of a high frequency board is not expressed by using a more expensive material alone. Material parameter stability, post-lamination dielectric thickness control, etch precision and impedance consistency all have to improve at the same time.

A manufacturer that can achieve the design value but cannot guarantee a stable distribution across production volume will be exposed when frequencies rise. The relevant evidence is measured impedance distribution across a panel and across lots, obtained through quality management systems that retain the data. Capability in PCB fabrication becomes meaningful only when that distribution is controlled rather than observed.

Dual Band Coexistence Expands the Problem

The harder challenge comes from frequency bands sharing one device.

A phone supporting both Sub-6GHz and sub-terahertz links has to manage coexistence across bands that differ by more than an order of magnitude in wavelength. Harmonic relationships, intermodulation products and blocking effects all have to be evaluated, and the board layout determines how much isolation is achievable between the relevant circuits.

Isolation is largely a layout property. It depends on physical separation, on the continuity of ground and reference planes, on the placement of filters relative to the aggressor source, and on the return path geometry of the currents that carry the interfering signal. None of those can be corrected by firmware or calibration once the board is fabricated.

Antenna placement and ground plane partitioning follow from the same constraints. Multiple antennas operating across widely separated bands require reference planes that serve each of them without creating coupling paths between them. On a handset where the antenna volume is already constrained by mechanical design, achieving that requires the RF engineer and the board designer to work from the same layout model rather than in sequence.

Filter design interacts with board parasitics as well. A filter specified to provide a given rejection at a given frequency will deliver that rejection only if the impedance it sees at its terminals matches what the design assumed. Via inductance, pad capacitance and trace discontinuities shift that impedance, and the shift becomes more significant as frequency rises. This is why PCB design and layout review for high frequency products has to treat component placement, reference plane design and via structure as part of the RF design rather than as a mechanical enclosure for it.

What Changes in the Manufacturing Specification

Translating the RF requirement into manufacturing terms produces a specific list of tightened specifications.

Dielectric thickness tolerance after lamination becomes a primary control parameter, because it sets the effective dielectric constant of the transmission structure and therefore the impedance and electrical length. Copper foil surface roughness is selected for the frequency range rather than for adhesion convenience. Etch precision on the trace sidewalls determines whether the impedance calculated in simulation is realized in copper.

Layer-to-layer registration matters more because via position determines the parasitic environment, and on a differential structure a small asymmetry converts differential signal into common mode noise. Surface finish selection becomes consequential, since at high frequencies the skin effect concentrates current in a thin layer whose conductivity depends on the finish chosen.

Each of these is a process specification rather than a design preference. The practical consequence for development teams is that the fabricator needs to be involved while the RF stackup is being defined, because the achievable tolerance window depends on the process being used and cannot be assumed from a generic material datasheet. For programs building high frequency prototype hardware, rapid PCBA prototyping provides the fastest way to measure real performance against simulation, provided the measurement methodology is held constant between iterations.

What Engineering Teams Should Do Now

Even with the regulatory timeline uncertain, several actions are worthwhile for teams developing RF hardware that may eventually face higher band requirements.

First, characterize the material system rather than assuming it. Measure dielectric constant and thickness variation across panels for the specific laminate and stackup being used, because the values that matter are the production distribution rather than the datasheet nominal.

Second, verify impedance on production coupons. A coupon structure fabricated on the same panel as the functional board provides a direct measurement of what the process produced, and it separates design error from manufacturing variation.

Third, design the reference plane architecture deliberately. Ground continuity beneath high frequency traces, avoidance of plane splits under critical nets and controlled via transitions are all decisions made during layout that cannot be revisited later at acceptable cost.

Fourth, plan for measurement in the prototype phase. Building test structures alongside functional boards allows RF performance to be evaluated directly rather than inferred from system behavior, and the resulting data is what supports an eventual certification claim.

Why Certification Timing Matters Less Than Design Timing

There is a reasonable question about whether it makes sense to design for requirements that are not yet formally mandatory. The answer lies in how long it takes to establish process capability.

Regulatory requirements typically arrive with a compliance date, and the industry response is to prepare before that date. But the preparation involved here is not documentation. It is the establishment of material databases, lamination parameters, drilling windows and impedance control limits for higher frequency operation. Each of those takes multiple build cycles to characterize, and characterization cannot be compressed at the end by working harder.

A program that begins this work when the requirement is published will be competing for the same fabrication capacity as every other program doing the same thing. A program that begins earlier encounters less contention and has more time to iterate on the design while changes are still inexpensive.

There is also a cost dimension. Designing with tighter tolerance than required increases board cost today. Designing with a stackup that cannot be upgraded to tighter tolerance later means a full redesign when the requirement arrives. The practical middle path is to choose materials and stackups that are current-production appropriate while confirming that they have a documented path to tighter control if needed, and to record the process capability data that would support that transition.

The Industry Signal

The useful conclusion from the sub-terahertz compliance discussion is not a date. It is that RF design constraints are entering existing product development flows earlier than the corresponding commercial networks.

Handsets are being designed today for spectrum that will not carry commercial traffic for years, and the manufacturing tolerances that support those designs have to be established before the designs are frozen. For telecommunications PCBA programs, that means the material, stackup and process decisions carry more weight relative to the electrical design than they did in previous generations, because the electrical design has less margin to absorb manufacturing variation.

Manufacturers that respond by building and documenting tighter process control, rather than by simply stocking higher grade materials, will be the ones able to serve the next generation of RF hardware. Frequency determines the requirement, but process capability determines whether the requirement can be met at volume.