A rigid-flex PCB copy is possible, and the copper geometry is rarely the problem. What decides whether the job succeeds is the stack-up and the transition zone — how many flex layers run through the bend, where the polyimide coverlay starts and stops, where the adhesive-free bond region sits, and how the rigid sections are drilled relative to the flex core. Get that wrong and you get a board that passes electrical test and then cracks after a few hundred flex cycles.
That’s the honest summary. The rest of this page explains what happens to your board in the lab, what we can and can’t recover, and what changes the answer.
Why rigid-flex is a different job from multilayer FR-4
On a conventional multilayer board, layer separation is destructive but predictable. You grind or chemically strip one layer at a time, image the copper, and the dielectric behaves the same way all the way across the panel. Rigid-flex doesn’t cooperate.
You have at least three different material regions in one part: rigid FR-4 sections, a polyimide flex core, and a transition zone where the rigid buildup stops and the flex continues. Each region delaminates differently. Polyimide is tough and doesn’t grind cleanly — it tears and stretches, which distorts the copper image you’re trying to measure. Coverlay bonds hard to the flex copper and often has to be removed chemically rather than mechanically.
Then there’s the part nobody advertises: the bond line. In a well-built rigid-flex the flex layers are deliberately unbonded through the bend region so the copper can slip and take the strain. That unbonded window has a defined length and position. It doesn’t show up in a copper image at all. You find it by cross-sectioning the transition, by X-ray, and by carefully peeling the actual sample. Miss it and your rebuilt board will be mechanically wrong even though every net checks out.
The three things that usually go undocumented
- Coverlay openings. Where the polyimide coverlay is windowed for pads, and whether those windows are laser-cut or punched. Tolerance here affects solder fillet and pad adhesion.
- Stiffener locations and material. FR-4 stiffener, polyimide stiffener, or a stainless plate? Thickness and adhesive type matter for connector insertion force.
- Bend radius intent. The original designer chose a minimum bend radius based on copper thickness and neutral-axis position. We can infer it from the built geometry, but if the assembly bends tighter than the board was designed for, the copy inherits the same fatigue life — no better.
How a rigid-flex PCB copy actually runs, step by step
Order matters, because several of these steps destroy the evidence for the ones after them. We do the non-destructive work first and completely.
- Intake and photography. Both sides, high resolution, calibrated against a scale. Every silkscreen mark, every date code, every rev letter gets captured before anything touches the board.
- Mechanical survey. Outline, flex arm lengths, bend positions, stiffener footprints, cutouts, tooling holes. On flex, the outline is functional — it’s often profiled to fit an enclosure path, so we measure it, we don’t approximate it.
- X-ray. Vias, buried structures, the drill pattern in the rigid sections, and — importantly — where copper stops in the transition. X-ray inspection of the assembled board also tells us whether any BGAs sit close enough to the flex tail to complicate rework.
- Component identification and BOM. Markings read under a microscope, plus in-circuit measurement for passives. Anything ambiguous gets flagged, not guessed.
- Desoldering. Low-temperature, controlled, with the flex supported. Polyimide tolerates heat better than FR-4, but the coverlay adhesive doesn’t like repeated cycles and the flex pads lift easily if you pull.
- Cross-sectioning a sacrificial coupon. Where a second board is available, we section the transition zone. This is the only reliable way to read the real stack-up: layer order, dielectric thicknesses, copper weights, and the extent of the unbonded region.
- Layer separation and copper imaging. Rigid sections first, then the flex core. Each layer is scanned, registered to a common datum, and traced.
- Rebuild and verification. Layers stacked in CAD, netlist extracted, then checked pin-to-pin against continuity readings taken from the original board before teardown. Discrepancies get resolved by going back to the scans, never by assuming.
The verification step is where this kind of job earns its keep. We compare three independent sources: the traced copper, the measured continuity from the intact sample, and the X-ray via map. When all three agree on a net, that net is settled. It’s the same discipline used in any careful rigid-flex PCB reverse engineering project, but here the flex layers are the ones most likely to disagree, because polyimide distortion shifts the image.
What we need from you before we can quote
Feasibility on rigid-flex is decided by looking, not by discussing. Send this and you’ll get a real answer instead of a range:
- Clear photos of both sides, whole board in frame, on a dark non-reflective background. Include a shot of the flex arms laid flat and one of the board in its installed bent position if you have it.
- Board dimensions — overall rigid section sizes, flex arm length and width, and total thickness at the rigid areas if you can measure it.
- Chip markings on the main controller and any custom or relabeled parts, typed out or photographed close up.
- Quantity available. One board, two, or more. This single fact changes the plan more than anything else.
- One line on deliverables: do you need Gerbers only, Gerbers plus BOM, a full schematic, or working boards back?
What comes back from us: a feasibility statement that names the specific risks on your board, a lead time, a price, and a note on which deliverables we’d consider high confidence versus best effort. If we think the flex layers can’t be recovered cleanly from a single sample, we say so up front. Price is driven by rigid layer count, flex layer count, the number of transition zones, component count, BGA presence, whether firmware is in scope, and which files you need — not by a flat rate.
How many boards you can supply changes everything
This is the constraint people underestimate. Layer separation is destructive. With one board, we have to choose between cross-sectioning the transition zone and keeping the board intact for continuity verification. You can’t do both to the same sample.
| Samples available | What’s realistic | Main risk |
|---|---|---|
| One board, non-functional | Copper geometry, BOM, Gerbers. Stack-up partly inferred. | Dielectric thicknesses and bond-line extent estimated rather than measured |
| One board, functional | Continuity map taken first, then destructive teardown | No sacrificial coupon for sectioning; nothing left to compare the rebuild against |
| Two or more boards | Full stack-up by cross-section plus an intact reference for verification | Mainly component identification on unmarked or relabeled parts |
| Two boards plus one you can flex-cycle | Everything above, plus mechanical validation of the rebuilt part | Cost and schedule, not technical uncertainty |
If you only have one unit and it’s the last spare for a machine still in service, tell us. There are non-destructive paths — X-ray plus surface tracing plus careful probing — that get you a usable schematic without cutting the board apart. They cost more per net and they won’t give you a certified stack-up, but the board goes back in the machine.
Where it gets genuinely hard, and where it stops
Cases that add time and risk:
- Fine-pitch BGA over a flex region. Rework support is awkward and the underlying copper is thin. Expect extra imaging passes.
- Blind and buried vias in the rigid sections. Every buried structure is another layer pair that has to be separated and registered independently, which is where controlled layer separation takes most of the schedule.
- Impedance-controlled pairs crossing the transition. The dielectric changes at the boundary, so the geometry changes to hold impedance. We measure the built dimensions, but reproducing the original impedance target depends on the fabricator matching the material set.
- Conductive adhesive or shielding film on the flex. Common on cameras and display tails. It has to be removed to see the copper, and it has to be specified correctly to restore EMI behavior.
- Potted or conformally coated assemblies. Coating removal on polyimide is slower and more delicate than on FR-4.
Cases where we may tell you no, or tell you the deliverable will be partial: badly cracked flex where copper is already broken and the original routing can’t be resolved; boards where the flex has been heat-damaged and the coverlay has bubbled over the traces; single-sample jobs where you need a guaranteed-equivalent stack-up for a qualification file. And separately from the board itself — if the design depends on a locked microcontroller, the copper copy gives you hardware but not behavior. Recovering the firmware is a separate scope with its own feasibility question, and no chip is guaranteed readable.
On rights: we work on the assumption that you hold the rights to the design or have authorization to have it reproduced, and every project runs under NDA. The usual legitimate reasons for this work are keeping out-of-production equipment running, recovering design data after files were lost, and dealing with a component that went end-of-life.
Acceptability criteria worth agreeing on up front
Before fabrication, decide which class you’re building to
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