RF PCB Copy: What Changes With Layer Count

On an RF board, layer count decides almost everything about an RF PCB copy job. Two-layer RF boards are mostly about recovering trace geometry and one dielectric thickness. Four layers and up means we have to reconstruct the stackup itself — core and prepreg thicknesses, which dielectric material was used, and which reference plane each controlled-impedance trace was designed against. Get that wrong and the copper looks perfect while the return loss doesn’t.

That’s the short version. Below is what actually happens on the bench, and where the honest limits are.

Why RF boards are a different job from digital boards

Copying a digital board is largely a connectivity problem. Copy every net correctly, keep the footprints right, and the board works. RF is a geometry-and-materials problem. A 50-ohm microstrip is 50 ohms because of trace width, copper thickness, dielectric height, and the dielectric constant of the substrate — change any one and the line moves off target.

So on an RF PCB copy we treat three things as first-class deliverables, not afterthoughts:

• Trace geometry to real dimensions, not just netlist connectivity. Line widths, gap widths on coupled lines, taper lengths, stub lengths, and radii on bends all get measured rather than redrawn by eye.
• The layer stackup, including the dielectric thickness under every controlled-impedance layer.
• The substrate material, or at least its electrical behavior — Dk and, where it matters, loss tangent.

Distributed structures make this unforgiving. Branch-line couplers, Wilkinson dividers, hairpin and interdigital filters, patch antennas, and quarter-wave matching sections are all defined by physical length in relation to wavelength in the medium. A 3-mil error on a 20-mil line is a real impedance shift. On a 2.4 GHz filter, a few mils of resonator length walks the passband.

What changes as you add layers

Two layers

Usually a single microstrip layer over one solid ground. This is the friendliest case. We image both sides, measure copper features under a microscope with calibrated optics, and cross-section a coupon from a non-functional edge to get board thickness and copper weight. One dielectric height to determine, one Dk to identify. Ground stitching vias are visible from the outside, so via placement around the RF path — which matters for mode suppression — is directly observable.

Four layers

Now the RF layer references an inner plane, and the plane may not be solid. RF designers cut voids under antenna feeds, split grounds around mixers, and sometimes reference layer 1 to layer 2 for RF while running DC and control between layers 3 and 4. We can’t see any of that from the outside, so the inner layers get imaged during controlled delamination and the plane cutouts are recorded as artwork, not summarized as “ground plane.”

The 2-to-4 jump is where turnaround stretches noticeably, for the same reasons that show up in a general side-by-side of two-layer and four-layer copy work — plus the extra impedance verification RF adds on top.

Six to twelve layers

Mixed-dielectric construction becomes common. A hybrid stackup might put a PTFE or ceramic-filled laminate on the outer RF layers and cheap FR-4 in the digital core, bonded with a bond ply that has its own Dk. Each interface has to be identified separately during cross-section. This is also where blind and buried vias, back-drilled through-vias, and cavity constructions turn up, and where X-ray becomes necessary rather than nice to have.

Above twelve layers

Usually a mixed-signal RF platform rather than a pure RF board — a radio front end sitting on top of a large digital backplane-style design. The RF portion is often only a handful of layers deep, but proving which layers are which, and confirming there are no buried structures you’d miss, takes real effort. Boards at this depth follow the same pattern described in our writeup on how single-layer and multilayer copies diverge, only with impedance targets layered on.

How we identify the substrate material

This is the question customers ask least and should ask first. Copper geometry is measurable. Dielectric constant is not printed on the board.

We work it from several directions at once. Legend markings and laminate identifiers are sometimes printed in a corner or etched into an unused copper area — free information when it’s there. Cross-section reveals glass weave style, resin content, and fill particles, which narrows the family fast: woven glass FR-4 looks nothing like a ceramic-filled hydrocarbon laminate under the microscope. Color and translucency after etching help. Thickness combinations often match a specific vendor’s standard core-and-prepreg menu.

Then we verify electrically. Many RF boards contain a structure whose behavior depends on Dk — a resonator, a coupled pair, a known-length line to a connector. Measuring the original board’s response and comparing it to a simulation lets us back out an effective Dk that reproduces the measured result. When the design has a test coupon or spare pads, TDR on the original gives us the line impedance directly.

Honest note: we recover an electrically equivalent material, not always the exact original part number. If your production requires a specific qualified laminate, tell us up front and we’ll flag which candidates match and where the substitution risk lies.

The process, in order

1. Intake and photography. Both sides documented before anything is touched, with a scale reference in frame.
2. Component identification and BOM. RF parts get careful treatment — SAW filters, baluns, VCOs, PA modules, and shielded crystals often carry house codes rather than catalog part numbers, and some are obsolete.
3. X-ray. Via structure, buried features, shield can contents, and any internal cavity get mapped before destructive work.
4. Scan and vectorize the outer layers. High-resolution imaging, then trace extraction to real dimensions.
5. Layer separation and inner-layer imaging. Controlled removal, layer by layer, with registration marks maintained so inner artwork aligns to the outer layers.
6. Cross-section. Dielectric heights, copper weights per layer, plating thickness in the barrels.
7. Stackup reconstruction and impedance math. Every controlled-impedance line gets calculated against its recovered stackup and compared to a sensible target — 50 ohms single-ended, 100 differential, whatever the topology implies.
8. Gerber, drill, and netlist output. Then a DRC pass and a netlist comparison against the original board.
9. Prototype and RF verification. First articles built on the reconstructed stackup, then measured — insertion loss, return loss, and passband against the original where a reference board is available.

Step 9 is the one that separates an RF PCB copy from a generic board duplication. A board that passes DRC and continuity can still be 3 dB down in the passband. We’d rather find that on our bench than yours.

What to send us, and what comes back

To quote an RF board we need very little, but the little we need has to be usable:

• Clear photos of both sides, on a dark non-reflective background, the whole board in frame, straight on rather than at an angle. Add close-ups of the RF section and of any shield can.
• Board dimensions and total thickness, measured with calipers if you have them.
• Layer count if you know it, and any laminate markings you can read on the board edge.
• Chip markings on the main RF parts — the transceiver, PA, or SoC — typed out, not just photographed.
• Operating frequency band and, if you know it, the impedance target. This one line saves us guessing.
• One line on deliverables: Gerber and drill only, or Gerber plus schematic, plus BOM, plus assembled samples.

What comes back from us is a feasibility assessment naming the specific risks on your board, a lead time, and a price. If we think an item is questionable — a house-marked SAW filter, an unreadable laminate, a suspected cavity — we say so before you commit, not after.

Physical samples matter more on RF than anywhere else. Two boards is much better than one: one goes to destructive analysis, one stays intact as the measurement reference. A single board means we have to choose between full stackup recovery and keeping a working comparison unit.

Where RF PCB copy gets hard, and where it sometimes isn’t feasible

Straight talk on the constraints:

Obsolete RF parts. A perfect board copy is useless if the transceiver went end-of-life six years ago. We flag these during BOM work and can propose alternates, but an RF front-end substitution frequently means a matching-network redesign, which is engineering work rather than duplication.

Potting and conformal coat over the RF section. Removable in many cases, but aggressive potting compounds can lift fine copper on the way out. On a filter with 6-mil resonator gaps, that’s a real risk to the very geometry we’re trying to measure.

Tuned-at-test designs. Some RF boards ship with trimmed traces, hand-placed solder blobs

Related reading

• Oct 2026ATSAMD51 Crack | Protection Levels & Recovery Guide
• Oct 2026Nuvoton IC Unlock | Families We Can Read & Recover
• Oct 2026ATSAMD51 Unlock | Feasibility, Cost & Lead Time