PCB replication is the process of taking a finished, populated board and rebuilding a complete manufacturing data set from it — Gerbers, drill files, stackup, BOM, and usually a schematic — so you can order new boards that behave like the original. You don’t need the original CAD files. You do need at least one physical sample, and ideally two if one can be sacrificed.
This page is written for the person who has a board in hand and no design data: a maintenance engineer keeping a line running, or a product owner whose original files walked out the door with a former contractor.
What actually happens to your board
The order matters, because each step depends on data captured in the one before it. Here is the sequence a typical replication job follows in our lab.
- Intake and documentation. The board is photographed at high resolution on both sides before anything is touched. We record dimensions, board thickness with a micrometer, copper weight where we can measure it, surface finish, and any date codes or version markings in the silkscreen.
- Component identification. Every part is logged by reference designator, package, marking, and measured value where the marking is ambiguous. Passives get measured off-board when tolerance matters. Unmarked or sanded parts get flagged here, not later.
- Electrical mapping. Continuity is traced net by net. On a two-layer board this is largely optical plus a meter. On a multilayer board we combine X-ray for via and inner-layer geometry with controlled delamination and layer-by-layer scanning.
- Layer imaging. Each copper layer is scanned separately at high DPI, registered against the drill pattern so the layers line up to the same origin. Registration error is what causes annular ring failures later, so this step gets checked twice.
- CAD rebuild. The scans become real vector geometry — pads, traces, planes, drills — not traced bitmaps. Trace widths and clearances are measured and entered as numbers, so the output can be checked against IPC-2221 spacing rules and the design rules of whichever fab you plan to use.
- Verification. The rebuilt netlist is compared against the netlist measured from the physical board, net by net. Then a first article is fabricated and, if the job includes assembly, populated and tested against the original.
The engineer’s attention through all of this is on the things that don’t photograph: which via is buried and which is through, whether that wide pour is a plane or a heavy-current trace, whether a resistor is 1% or 5%, and where the original designer deliberately controlled impedance. A 100-ohm differential pair copied at the wrong trace-to-trace spacing will pass continuity and fail at speed.
Two-layer versus multilayer
A two-layer board is close to a scanning and drafting exercise. Both copper layers are visible, and a careful engineer can verify most of it with a meter in an afternoon.
Four layers and up is a different job. Inner layers have to be imaged separately, which means the board is consumed, and plane splits are easy to get subtly wrong. HDI boards with stacked microvias and fine-pitch BGAs are the slowest cases, because via-in-pad structures and blind/buried transitions have to be resolved with X-ray before the stackup can be trusted. That’s where most of the schedule goes, not in drawing traces.
What you get at the end of a PCB replication project
Deliverables are tiered, and the tier you need depends on what you plan to do next. Buying only Gerbers is fine if you just want more of the same board. If you plan to change anything, you want the schematic too.
| Deliverable set | What’s included | Choose it when |
|---|---|---|
| Manufacturing data only | Gerber RS-274X, NC drill, stackup drawing, fab notes | You need identical replacement boards, no changes |
| Manufacturing + assembly | Above, plus BOM with references and pick-and-place file | You’re ordering assembled boards from a CM |
| Full engineering package | Above, plus editable CAD project, netlist, and schematic | You need to substitute parts, shrink the board, or fix a known weakness |
Most customers underestimate the value of the schematic until an obsolete part forces a change. If there’s any chance of a redesign in the next few years, get it done while the board is on the bench. Our page on what files come back from a copy job lists the formats we hand over.
What to send us before we can quote
We can give a real feasibility answer and a real number from photos alone in most cases. Send this, and skip the back-and-forth:
- Clear photos of both sides, whole board in frame, on a dark non-reflective background, well lit, in focus. Add close-ups of any dense area.
- Board dimensions — length, width, and thickness if you can measure it — plus layer count if you know it.
- Markings on the main ICs, especially the microcontroller, FPGA, or any custom part. A legible close-up beats a typed guess.
- Quantity and intent — five spares, or a production run.
- One line on deliverables — Gerbers only, or Gerbers plus schematic plus BOM.
- Whether firmware is involved, and whether you have a programmed sample and a hex file.
What comes back from us is a feasibility assessment naming the specific risks on your board, a lead time, and a price. If something on the board is questionable, we say so in the quote rather than after you’ve paid. Sending a physical sample for evaluation of an actual board is the fastest route when photos leave real doubt.
Where PCB replication gets hard, and where it stops
Any lab that tells you every board is easy hasn’t done enough of them. The honest list:
Potted and conformally coated boards. Hard epoxy potting over a dense assembly can destroy traces during removal. Sometimes the potting comes off cleanly with solvent and heat. Sometimes it takes the silkscreen and thin traces with it. We assess this before quoting and tell you the risk to your sample.
Sanded, relabeled, or house-numbered parts. A part marked only with a customer code has no public datasheet. We can often identify it from package, pinout, and in-circuit behavior, but “often” is not “always.” A custom ASIC or a mask-ROM part with no equivalent may have no path forward at all.
Single sample, multilayer. Layer separation is destructive. If you have exactly one board and it’s six layers, you’re choosing between keeping it and replicating it. X-ray plus surface scanning gets us part of the way, but inner-layer plane splits on a dense board are much less certain without delamination.
Locked microcontrollers. Copying the copper doesn’t give you the code. If the firmware is in a protected MCU, that’s a separate job with its own feasibility question — and the answer depends heavily on the part. An older 8051 with a simple lock bit is a different situation from an STM32 sitting at RDP Level 2, which is designed to permanently close the debug interface. We assess per part number and never promise a specific device is readable.
RF and controlled-impedance boards. Geometry can be copied precisely; dielectric constant cannot be read off a scan. Getting a 2.4 GHz front end or a high-speed backplane to perform identically means matching the laminate, not just the artwork. Expect material identification work and expect us to ask what performance you actually need.
Damaged samples. A board with burned traces, a blown-out area, or lifted pads has missing information. We reconstruct from the surrounding circuit and from the schematic logic, and we mark every reconstructed net as an assumption in the handover documents.
When replication is the right call — and when it isn’t
Replication makes sense when the original design works, the volume is modest, and the alternative is losing the equipment. Legacy industrial controllers, out-of-production drives, test instruments, and obsolete power supply boards are the classic cases. The design is proven, the qualification is already done, and a redesign would cost more than the boards are worth.
It’s the wrong call when a third or more of the BOM is obsolete, when the board was already marginal thermally or electrically, or when you’re about to change the product’s requirements anyway. At that point you want the schematic and a fresh layout, not a copy of a dead-end design.
Legitimate uses cover most of the work we see: maintaining equipment you own, recovering data for a design your company paid for, second-sourcing a part that went end-of-life, and failure analysis. You’re responsible for holding the rights to the design you send us; every project runs under NDA and your files never leave the lab.
Moving from a verified sample to volume
The first article is the real deliverable, not the Gerbers. We fabricate one, check it against IPC-A-600 and IPC-6012 acceptability criteria for plating, annular ring, and registration, then assemble and compare it to your original — same current draw, same signal behavior at the points you care about, same functional pass.
Only after that does the data set get released for volume. If you’re going straight into production, our notes on taking a verified sample into mass production cover the DFM and sourcing decisions that come next.
Frequently asked questions
Can you replicate a PCB without the schematic or any original files?
Yes — that’s the normal starting point. We work from the physical board. The schematic is an output of the process, not an input. What we can’t work around is a missing board: photos alone let us quote, but rebuilding manufacturable data requires the hardware on the bench.
Will you destroy my board?
Two-layer boards usually survive. Multilayer boards generally don’t, because inner layers have to be delaminated and imaged individually. If you can only provide one sample and it must come back intact, tell us up front — we’ll quote an X-ray-based approach and be explicit about which parts of the stackup carry more uncertainty.
How long does PCB replication take?
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