On a high frequency PCB copy, layer count matters less for the copper geometry than for the dielectric. Two-layer RF boards on a single laminate are usually straightforward: identify the material, measure the copper, match the impedance. Once you get to four, eight or twelve layers with mixed dielectrics, blind vias and buried reference planes, the hard part becomes reconstructing the stackup — because the trace widths in the Gerber mean nothing without the exact substrate under them.
That’s the short answer. Below is what actually happens at each level, and what changes in price, lead time and risk.
Why a high frequency board can’t be copied like a digital board
Copy a 2-layer control board wrong by half a mil of trace width and nothing happens. Copy a 50 Ω microstrip half a mil narrow on a 0.508 mm PTFE laminate and you’ve moved the impedance by a couple of ohms. Do that on a matching network at 5.8 GHz and the return loss falls apart.
So a high frequency PCB copy is really three jobs stacked together:
- Geometry recovery — trace widths, gaps, ground clearances, via barrel and pad sizes, stub lengths, taper shapes. On RF boards the copper shapes are the components. A quarter-wave stub is a part, not a routing artifact.
- Material identification — which laminate, which Dk, which copper foil roughness, what surface finish. This is where most cheap copies fail silently.
- Stackup reconstruction — dielectric thickness between each signal layer and its reference plane, prepreg vs core, and where the reference plane actually is when planes are split.
The first job scales gently with layer count. The second and third scale steeply.
What changes at 2, 4, 6-8, and 10+ layers
Two layers
Almost always a single core with microstrip on top and a solid ground below. One dielectric to identify, one impedance case to solve. We measure overall board thickness, subtract two copper weights and the finish, and we have the dielectric thickness within a reasonable band. Then we solve for Dk using the observed trace width against the known target impedance of the interfaces on the board — a 50 Ω SMA launch is a very useful piece of evidence.
These are the boards where a good copy behaves like the original on the first prototype. Amplifier boards, simple filters, antenna feeds, ISM-band modules.
Four layers
The common RF layout is signal / ground / power / signal, and the RF stays on the outer layers with a tight ground beneath. The added work is that inner layers must be imaged separately after controlled delamination, and the two dielectric thicknesses are usually different — a thin RF core bonded to a thicker FR-4 sub-assembly is very common in hybrid builds.
Hybrid stackups are the thing to watch here. A board can be Rogers 4350B on the top pair and standard FR-4 below, which means the copy house has to notice the color and texture change on the cross-section instead of assuming one material throughout. If you’ve read our comparison of 2 layer versus 4 layer copy work, the same logic applies, just with material identification bolted on top.
Six to eight layers
Now you get buried reference planes, split planes, and stripline instead of microstrip. Stripline impedance depends on the dielectric above and below the trace, so an error in either thickness moves the result. Plane splits matter enormously: a differential pair that crosses a split has a return path problem the original designer either handled deliberately or got away with, and copying it faithfully means copying that choice too.
Eight-layer high frequency boards are also where via structures stop being simple. Back-drilled vias, via stubs deliberately left or removed, and ground stitching patterns around signal transitions all have to be measured, not guessed. X-ray helps a lot here for hole depth and internal via structure before any destructive work starts.
Ten layers and up
At this point the board is usually a mixed-signal or backplane-class design and the RF or high-speed portion is one region of a bigger board. Expect blind and buried vias, sequential lamination, possibly HDI microvias in a BGA field. Layer separation has to be done in sequence and documented photographically at every stage, because you only get one pass on a given sample.
Realistically: a 12-layer high frequency board with a fine-pitch BGA is a different class of project from a 2-layer filter board. Different price bracket, different lead time, and we’ll want more than one sample.
| Layer count | Main technical driver | Typical extra work | Risk of a functional miss |
|---|---|---|---|
| 2 | Single dielectric, microstrip | Material ID, impedance solve | Low |
| 4 | Possible hybrid stackup | Delamination, two dielectric measurements | Low to moderate |
| 6–8 | Stripline, split planes | Per-layer imaging, via structure mapping | Moderate |
| 10–16 | Sequential lamination, blind/buried vias | X-ray, staged separation, multiple samples | Higher — needs test verification |
| 18+ | Backplane-class, very thick stack | Full stackup documentation, impedance coupons | Highest — plan a validation build |
How we identify the laminate without the original datasheet
Nobody prints “Rogers RO4003C, 0.508 mm” on the silkscreen. So we work from evidence:
- Cross-section. A polished cross-section under the microscope gives dielectric thicknesses layer by layer, copper weights, plating thickness in the barrels, and whether a layer is core or prepreg.
- Visual and physical character. PTFE-based laminates, ceramic-filled hydrocarbons and standard high-Tg FR-4 look, cut and behave differently. Color, glass weave visibility, and how the material responds to routing all narrow the field.
- Back-solving from known impedances. Where the board has a connector or a device with a defined port impedance, the trace geometry plus the measured dielectric thickness leaves Dk as the only unknown. Solve it, then check that the answer is consistent across other traces on the same layer.
- TDR on the original. If the sample is intact and has usable launch points, a time-domain reflectometry measurement on the real board is the most direct evidence available, and it’s what we check the rebuilt board against later.
- Sanity check against what’s buildable. There’s a finite set of laminates a fabricator can actually source. We pick the closest available match and state the substitution explicitly rather than quietly using FR-4 and hoping.
That last point is the honest one. Sometimes the original material is discontinued or single-sourced, and the correct deliverable is “here is the closest current equivalent, here is the Dk difference, here is the recommended trace width adjustment.” That’s a design change, and you should be told about it in writing. It’s the same discipline we apply on impedance controlled board work generally, and it’s closely related to how we approach a dedicated RF board duplication project.
What to send us, and what comes back
To quote a high frequency PCB copy accurately, send:
- Clear photos of both sides, whole board in frame, on a dark matte background, with even lighting and no flash glare on the solder mask
- Board outline dimensions and overall thickness (a caliper reading is fine)
- Layer count if you know it, or just say “unknown”
- Close-ups of the main IC markings — RF front ends, PLLs, transceivers, anything with a laser-etched part number
- The operating frequency range, if you know it, and any interface impedance (50 Ω, 75 Ω, 100 Ω differential)
- One line on what you need back: Gerbers only, Gerbers plus stackup and impedance report, full schematic, BOM, or working assembled boards
- Whether the sample is expendable — cross-sectioning is destructive
What you get in return before any money moves: a feasibility read, a lead time, a price, and a note on anything we think is a risk on your specific board. What you get at delivery: Gerbers and drill files, a stackup drawing with dielectric materials and thicknesses called out, an impedance table for the controlled nets, the BOM, and — if you asked for it — assembled boards with the RF path verified against the original sample.
Where high frequency board duplication gets hard, or doesn’t work
Straight talk on the limits:
- Discontinued laminates. If the original substrate is no longer made, an exact copy isn’t possible. A functional equivalent usually is, with adjusted geometry.
- Absorbers, embedded passives and buried resistive films. These show up on some microwave assemblies and are not recoverable by ordinary imaging. They need to be identified and specified deliberately.
- Tuned-by-hand boards. Some production RF boards were trimmed at test — a shortened stub, a scraped ground pad, a hand-placed capacitor. Copying the artwork copies the pre-tuning state, and your build will need its own tuning pass.
- Single damaged sample. Heavily corroded, delaminated or burned boards lose exactly the information we need. Cross-sectioning a burned area tells you nothing useful.
- Antennas and structures that interact with the enclosure. The board copy can be perfect and the performance still different if the mechanical environment isn’t reproduced.
None of these are dealbreakers by default. They’re things to know before you commit, and the reason our first response is a feasibility read rather than a price list.
On rights: we work on boards you own or are authorized to reproduce — obsolete equipment you maintain
Working on a board like this?
Send the chip marking or two photos. You get feasibility, lead time and price within 24 hours, and the check costs nothing.
Get a free quote