Contact Us
  • Home
  • BLOG
  • Why Does Copper Foil Roughness Matter Above 20 GHz?

Why Does Copper Foil Roughness Matter Above 20 GHz?

Because above 20 GHz the current stops flowing through the copper and starts flowing along the top 0.46 um of it, while the tooth profile of standard electrodeposited foil is 5 to 8 um deep. The current has to climb every tooth, its path is longer than the straight line between two points, and resistance and insertion loss rise with it. On a 20 GHz channel that effect is larger than most of the stackup arguments that reach a quotation, and it is usually the one item missing from the drawing.

Skin depth is the number that decides it. In copper it falls from 2.06 um at 1 GHz to 0.65 um at 10 GHz, 0.46 um at 20 GHz, 0.33 um at 40 GHz and 0.24 um at 77 GHz. Roughness is not measured against zero, it is measured against that depth, and the ratio of Rz to skin depth is what sets the penalty. The same foil that is harmless at 1 GHz is a different conductor at 40 GHz.

FrequencySkin depth in copperRz 6 um foil, ratioRz 2 um foil, ratio
1 GHz2.06 um2.91.0
10 GHz0.65 um9.23.1
20 GHz0.46 um13.04.3
40 GHz0.33 um18.46.1
77 GHz0.24 um25.58.5

What the ratio does to loss

The standard correction, the one most field solvers use, is the Hammerstad factor: the resistance multiplier rises with the square of Rz over skin depth and saturates at about 2. In practical terms the roughness penalty has a ceiling of roughly double the conductor resistance, and standard foil reaches that ceiling somewhere between 10 and 20 GHz. HVLP foil at Rz 1.5 to 2 um does not reach it below 40 GHz. That gap is the entire commercial reason the smooth foils exist, and it is why the gain from changing foil class is easy to measure and the gain from changing Dk by 0.1 in the same stack usually is not.

Foil classes are described by exactly that profile: standard ED foil at Rz 5 to 8 um, VLP at 2 to 3 um, HVLP at 1.5 to 2 um. When a stackup says "low loss" without naming the foil, the sentence describes the laminate and says nothing about the copper, which is half of the loss at these frequencies.

The trade you make when you go smooth

Peel strength comes from mechanical keying, so a smoother foil bonds less well to the same resin. IPC-6012 asks for about 1 N/mm, roughly 5.7 lb/in, on 1 oz copper, and it asks for it after thermal stress. Smooth foil meets that number only with a chemical bonding treatment, an oxide alternative or a thin passivation, and the treatment is where the two requirements meet. If a supplier pairs HVLP foil with a standard brown oxide, one of two things happens: peel drops after the 288 C float, or the treatment is thick enough to restore the roughness the smooth foil was bought to remove. Ask which treatment is used, and ask for peel after float rather than peel as received.

The same logic applies to the plated copper on top. A 20 to 25 um plating over rough base copper keeps the rough interface at the base, and the plating adds its own surface. In an RF stack the conductor is the foil, the plating and the finish in series, and the smoothest of the three sets no limit on the other two.

Where dielectric loss stops being the whole story

Total insertion loss splits into conductor and dielectric. On FR-4 at Df 0.02 the dielectric dominates at 20 GHz and the foil choice is second order. On RO4003C at Df 0.0027 the two are comparable. On PTFE at Df 0.0009 to 0.002 the conductor dominates above about 30 GHz, and foil roughness, plated copper and surface finish become the variables that decide the result. The finish belongs in the same conversation: 3 to 6 um of nickel phosphorus in ENIG is a resistive layer sitting in the current path, which is why immersion silver or bare copper with an OSP is the usual choice above 20 GHz.

Three mistakes that repeat

Choosing the laminate by Dk alone. A stack can be moved from FR-4 to a hydrocarbon laminate and still lose more than expected because the foil was left at standard profile and the finish was left at ENIG. Two changes on the copper side usually matter more than the third decimal of Dk.

Measuring at 1 GHz and extrapolating. The Rz over skin depth ratio moves by a factor of twenty between 1 GHz and 77 GHz, so a coupon measured at 1 GHz cannot rank two foils. Coupon data has to be taken in the band the product uses, with the same line width and the same finish.

Forgetting the return path. Roughness on the signal layer is only half the loop. The reference plane carries the return current, and a plane with the same rough foil adds the same multiplier. Symmetric stacks with smooth foil on the signal layers and standard foil on the planes give about half the benefit people expect.

What SUPERB PCB controls on an RF build

500 MHz to 77 GHz, Rogers RO4003C, RO4350B and RO4835, Taconic RF-35, TLY-5 and CER-10, PTFE with plasma treatment before plating, RF and FR-4 hybrid stacks, 0.05 mm line and space, impedance to plus or minus 5 percent with plus or minus 2 percent on critical nets, back drill stub of 0.25 mm or less with depth control of 0.05 mm, board thickness 0.2 to 12 mm, up to 68 layers, and a TDR coupon on every panel.

The RF side is on RF PCB and RF and microwave PCB, the material comparison is in Rogers, PTFE or FR-4, and the finish question is in which surface finish above 10 GHz. SUPERB Tech quotes the foil class in writing on the stackup, because the copper is the part of a low loss stack that is easiest to substitute quietly.

What to send with the question

Frequency band, layer count and stack, the loss budget in dB per inch, the foil class you will accept, the bonding treatment, the finish, line width and the dielectric thickness above the plane, and any coupon data you already have in band. With those seven items a quotation can be built against a loss target instead of against a list of materials, and the difference between the two is usually where a 40 GHz board either meets its budget or does not.

Need a quote or a review of your stack against these limits? pcba@superb-tech.com · WhatsApp +86 133 9608 1443