mmWave Copper Roughness: Loss, Dk Shift & Foil Guide

Why mmWave Copper Roughness Dominates Loss

At millimeter-wave frequencies (30–300 GHz), the skin depth in copper shrinks below 0.4 µm, forcing current to follow every peak and valley of the foil surface. Rough copper can add 0.5–3 dB/inch of extra conductor loss compared with a perfectly smooth conductor, and it raises the effective dielectric constant (Dk_eff) of a microstrip line by 2–8 %. These two effects—excess loss and Dk shift—are the primary reasons why mmWave copper roughness matters more than laminate Dk tolerance in many 60–79 GHz designs.

Understanding the interaction between roughness parameters, foil grades, surface finishes, and electromagnetic models is essential for closing the gap between simulation and measurement. The sections below walk through the physics, the numbers, and the practical choices that let you control copper roughness at mmWave frequencies.

Skin Depth and the Roughness Scale Problem

Skin depth δ in copper at a given frequency is:

δ = 1 / √(π f μ₀ σ)

At 10 GHz, δ ≈ 0.66 µm. At 77 GHz (automotive radar), δ ≈ 0.24 µm. Standard electrodeposited (ED) copper foil has an Rz (mean peak-to-valley height) of 3–6 µm on the matte side—roughly 12–25 × the skin depth at 77 GHz. Current cannot “average out” those features; it must traverse a longer, lossy path. The ratio Rz/δ is the single most useful indicator of how much mmWave copper roughness will hurt a design.

When Rz/δ exceeds roughly 2, roughness correction factors become significant. Above 10, they dominate the conductor loss term. This is why designs at 5G FR2 bands and automotive radar cannot rely on standard copper foil and still hit their link budgets. For a deeper look at how Rz, Rq, and Ra relate to each other and to vendor datasheets, see the guide to copper roughness Rz, Rq, and Ra specifications.

How mmWave Copper Roughness Increases Conductor Loss

Classical Roughness Correction Models

Several models correct the smooth-conductor attenuation constant α_c by a multiplicative factor K_r:

• Hammerstad–Jensen (1980): K_r = 1 + (2/π) arctan(1.4 (Δ/δ)²), where Δ is the RMS roughness. Simple and widely available in field solvers, but it saturates at K_r = 2 (a 2× loss increase), which underestimates measured loss above about 40 GHz.
• Huray snowball model: Treats the surface as a lattice of hemispherical protrusions. It fits measured data much better at mmWave because it accounts for the fractal-like nature of real foil. Requires the number of spheres per unit area (N) and sphere radius (r), parameters that foil vendors are increasingly publishing.
• Gradient model (Bracken, Triverio): Replaces the rough surface with a graded-conductivity transition layer. Good accuracy with fewer parameters, gaining traction in commercial 2.5-D solvers.

At 77 GHz on standard ED foil, the Huray model predicts K_r ≈ 2.5–3.0, meaning the conductor loss is 2.5–3 × higher than smooth-copper calculations predict. On HVLP (hyper-very-low-profile) foil the same model gives K_r ≈ 1.4–1.7. That difference can easily be 1–2 dB/inch on a 50 Ω microstrip—enough to wreck a 77 GHz radar front-end’s noise figure if not accounted for.

When building an RF simulation workflow, always confirm which roughness model your solver uses and feed it measured foil parameters rather than defaults. A Hammerstad model with default Δ will under-predict loss at mmWave by 30–50 %.

Loss Budget Impact

Consider a 77 GHz microstrip feed network on Rogers RO3003G2 (Dk 3.00, Df 0.0010 at 10 GHz). On a 2-inch trace:

The difference between standard ED and HVLP foil across just two inches is 2.2 dB—more than many mmWave amplifier stages provide in gain. This underlines why foil grade is a first-order design parameter, not a procurement afterthought. For a complete accounting of where loss accumulates from die to antenna, see the mmWave transition loss budget breakdown.

How Roughness Shifts Effective Dk

Roughness does not only add loss. The extended current path along rough surfaces increases the effective electrical length of the conductor, which a field solver or TDR measurement interprets as a higher Dk_eff. Published studies (e.g., IEEE EPEPS 2016 papers on roughness-induced Dk shift) show shifts of 3–8 % when moving from HVLP to standard ED foil on the same laminate, at frequencies above 30 GHz.

This matters because:

• Impedance error: A 5 % Dk increase shifts a 50 Ω microstrip to roughly 48 Ω—borderline acceptable at 28 GHz, unacceptable for a tightly matched 77 GHz patch feed.
• Phase error: In phased-array feed networks, a 5 % Dk shift produces ~2.5 % phase-length error per element, enough to degrade sidelobe levels by several dB.
• Correlation gap: If your EM solver assumes smooth copper, or uses the wrong roughness model, your simulated S-parameters will not match VNA measurements. Closing that gap is a core part of any de-embedding and correlation workflow.

Because roughness-induced Dk shift is frequency-dependent—larger at higher frequencies where δ is smaller—it compounds the natural Dk and Df frequency dependence of the laminate itself. Always extract Dk from measurements at the operating frequency, on the actual foil grade, rather than trusting 10 GHz datasheet values.

Copper Foil Grades for Controlling mmWave Copper Roughness

Foil vendors classify roughness into grades. The naming is not fully standardized, but the industry has converged on these tiers:

The trade-off is always roughness versus adhesion. Smoother foil bonds less mechanically to the dielectric, which can cause delamination under thermal cycling or during assembly reflow. Laminate vendors address this with coupling agents and oxide-alternative treatments, but the designer must verify peel strength meets IPC-6012 Class 3 requirements for the intended reliability class.

Use the tool below to narrow down which foil grade fits your frequency, loss budget, and reliability requirements.

If the selector recommends HVLP or EULP foil and your laminate vendor only qualifies VLP on that resin system, you have a conversation to start with your fabricator early—not after prototypes fail thermal shock testing.

Surface Finish Effects on mmWave Copper Roughness and Loss

Copper roughness is only half the conductor-surface story. The surface finish applied to exposed pads and transmission-line features adds its own loss and can modify roughness.

For mmWave transmission lines, immersion silver or OSP are the preferred finishes. ENIG is common in mixed-signal boards but should be kept off RF signal traces above ~20 GHz. If your board requires ENIG for BGA assembly on digital sections, specify selective finishing: ENIG on BGA pads, immersion silver on RF traces. This adds fabrication cost and process steps, so discuss it with your fab during the surface finish selection stage.

Answer a few questions about your design and the quiz below will recommend the best surface finish for your frequency and assembly constraints.

If the quiz recommends immersion silver but your assembly house flags shelf-life concerns, consider specifying a nitrogen-purge dry-pack requirement in your fabrication notes and plan assembly within 48 hours of opening.

Modeling mmWave Copper Roughness in EM Simulators

Getting the Parameters Right

Most commercial 3-D and 2.5-D solvers (HFSS, CST, Momentum, Sonnet) now support at least Hammerstad and Huray roughness models. To use them effectively:

1. Obtain foil-specific data. Ask the laminate vendor for Huray parameters (sphere radius and tile ratio/density) for the exact foil grade laminated to your chosen resin. Rogers publishes roughness data for several foil options on their high-frequency laminates.
2. Assign roughness to both sides. The drum side (smooth) and matte/treated side of the foil have different Rz values. In a microstrip, the side facing the dielectric (matte/treated) dominates because that is where the highest current density flows near the ground reference.
3. Validate with measurements. Fabricate a set of calibration structures (through lines, open stubs) and extract attenuation and Dk_eff versus frequency. Compare against simulation. If your correlation is off by more than 0.1 dB/inch or 2 % in Dk at the operating frequency, revisit your roughness parameters.
4. Use Touchstone files for system simulation. Once your EM model is correlated, export S-parameter data as Touchstone .s2p files for use in circuit-level co-simulation.

Common Pitfalls

• Using Rq when the model expects Δ or Rz. These are different statistical measures. Rz ≈ 4–6 × Ra for typical ED foil; Rq ≈ 1.1–1.3 × Ra. Mixing them up produces wildly wrong corrections.
• Ignoring oxide/roughening treatments. Some fabricators apply brown or black oxide to inner layers for adhesion. This adds ~0.5–1.0 µm to the effective roughness and should be included in the model.
• Assuming both conductor surfaces are identical. Electrodeposited foil is asymmetric. Rolled-annealed (RA) foil is smoother and more symmetric but is less common on RF laminates.

Design Guidelines for Minimizing Roughness Impact

1. Specify HVLP or better foil for any design operating above 30 GHz. Below 10 GHz, standard foil is usually acceptable; between 10–30 GHz, VLP is a reasonable compromise.
2. Choose immersion silver or OSP on RF signal traces. Reserve ENIG for non-RF pads only.
3. Minimize trace length. Roughness loss scales linearly with length. Use substrate-integrated waveguide (SIW) or grounded coplanar waveguide (GCPW) to shorten runs where possible.
4. Use stripline for long runs. Both reference planes constrain fields, reducing radiation loss—but both copper interfaces contribute conductor loss. Weigh this trade-off explicitly.
5. Request foil-specific Dk/Df data at your operating frequency from the laminate vendor. Do not design to 10 GHz datasheet numbers for a 77 GHz product.
6. Verify peel strength. Specify the IPC-TM-650 2.4.8 test and confirm the fabricator can meet your reliability tier with the chosen smooth foil.
7. Include loss margin. Even with HVLP foil and good modeling, fabrication lot-to-lot Rz variation of ±0.2 µm translates to ±0.1–0.2 dB/inch uncertainty at 77 GHz. Budget for it.

When selecting your high-frequency laminate, remember that the foil grade shipped on a given product code can vary by region and date. Always confirm the specific foil option in your purchase order.

Frequently Asked Questions

Does copper roughness affect Df as well as Dk?

Roughness increases the measured effective loss tangent because the additional conductor loss is mathematically indistinguishable from dielectric loss in a simple extraction. True bulk Df of the resin does not change, but any extraction that lumps all loss into Df will overstate it if roughness is not modeled separately.

Can I use FR-4 with smooth copper for mmWave?

No. Even with HVLP foil, FR-4’s high Df (0.015–0.025 at 10 GHz, rising further at mmWave) makes dielectric loss prohibitive above ~10 GHz. Smooth foil helps, but it cannot compensate for a lossy resin system.

How do I measure copper roughness on a finished board?

Cross-section microscopy gives Rz directly. Non-destructive options include white-light interferometry and atomic force microscopy (AFM) on a coupon. Specify roughness measurement coupons in your panel design if roughness control is critical.

Is rolled-annealed (RA) foil better than ED for mmWave?

RA foil is inherently smoother (Rz ~0.3–0.6 µm) and works well for flex circuits. However, RA foil is not commonly laminated to rigid RF substrates like RO4350B or RO3003. Availability and adhesion must be confirmed with the laminate supplier before specifying it.

What roughness model should I use in HFSS?

For frequencies above 20 GHz, the Huray model with vendor-supplied sphere radius and tile ratio gives the best correlation. The built-in Groiss (modified Hammerstad) model is acceptable below 20 GHz but under-predicts loss at mmWave.