An LED Driver PCB Copy takes a working or failed driver board and turns it back into manufacturable data: Gerbers, drill files, a BOM with real part numbers, and a schematic if you want one. For a two-layer isolated flyback driver with a coated top side, that’s usually a matter of days. Where it gets interesting is the magnetics, the potting compound, and the primary-side controller — those three things drive most of the cost and most of the honest caveats below.
Why LED driver boards are their own category
A driver board isn’t a logic board with a power section bolted on. The whole thing is a switching converter, and the copy has to respect that. Creepage and clearance between primary and secondary aren’t decoration — if the redrawn layout narrows the isolation gap under the optocoupler or across the transformer footprint, the board fails a hipot test and possibly a safety review.
So the geometry gets treated as functional data, not artwork. We measure the actual primary-to-secondary slot width, the pad-to-pad spacing across the Y-capacitor, the copper area used as a heatsink for the MOSFET, and the star point where the current-sense return ties back to the controller ground. IPC-2221 gives the general spacing framework, but on a mains-input driver the original designer usually spaced things to a safety standard the equipment was certified against, and that spacing has to survive the copy untouched.
The parts that hide their values
Three component classes cause almost all of the guesswork on LED driver boards:
- The transformer or inductor. Usually a custom wound part with a house number and nothing else. Turns ratio, core material, gap, and winding order all matter to the output regulation and the EMI signature.
- The controller IC. Frequently laser-marked with a house code, sometimes sanded, sometimes a die in a SOP-8 that a hundred vendors also sell.
- Current-sense resistors. Low value, often four-digit codes or unmarked, and the LED current depends directly on them. Getting the milliohm value wrong by 20 percent means the fixture runs 20 percent off spec.
For the magnetics we reverse the part when the customer needs a buildable BOM: measure primary and secondary inductance, ring the turns ratio, note wire gauge and layering, and — if the customer authorizes it and we have a spare unit — physically unwind one to confirm. That destroys the sample, so we always ask first. For sense resistors we measure with a four-wire meter rather than trusting the marking.
What actually happens to your board, in order
- Intake and photography. Both sides shot at high resolution, plus close-ups of every marking. This becomes the reference we check the finished data against.
- Conformal coating and potting removal. Most outdoor and high-humidity drivers are coated; many are fully potted. Coating comes off with solvent and controlled heat. Potting is a slower job and sometimes the deciding factor on feasibility.
- Component identification. Values measured in circuit where possible, out of circuit where not. House-marked semiconductors get identified by pinout, package, and circuit function — a primary controller with a HV startup pin, a COMP pin, and a CS pin narrows down fast.
- Desolder and clean. Board stripped, pads cleaned, silkscreen and reference designators recorded before anything gets abraded.
- Layer imaging. Two-layer boards scan straight through. Four-layer and up need the same layer-by-layer separation we use on motor control boards, where inner planes and buried traces have to be imaged one at a time.
- Vectorization and rebuild. Copper traced into a real CAD file with pad stacks, drill sizes, plane cutouts, and the original spacing preserved. Not a bitmap traced by software and shipped.
- Verification. Netlist from the new CAD data compared against continuity measured on the physical board, net by net. Board outline, mounting holes, and connector positions checked with calipers against the original.
- Optional build and test. Bare boards fabbed, assembled, and run up on a load — output current at nominal and worst-case line, ripple, thermal soak on the switch and diode.
The verification step is the one that separates useful data from a pretty picture. Any driver copy that hasn’t been continuity-checked against the source board will have at least one wrong net, and on a switching supply a wrong net near the gate drive or the feedback loop is not a cosmetic problem.
What you get, and how deep you want to go
| Deliverable set | What’s included | Good for |
|---|---|---|
| Fabrication data only | Gerbers, drill, stackup notes, board outline, silkscreen | You just need more bare boards of a driver that already works |
| Fab data + BOM | Above, plus BOM with measured values, package sizes, and sourcing notes on custom parts | You need a full build and the original supplier is gone |
| Full reverse engineering | Above, plus editable schematic, netlist, magnetics spec, and design intent notes | You need to modify output current, change the LED string, or re-certify |
| Copy + firmware | Above, plus code read from a dimming or DALI/0-10V control MCU where feasible | Digitally dimmable drivers where the source code is lost |
Most requests that start as “just give me the Gerbers” end up wanting the schematic once the customer realizes they want to change something. If you plan to re-spec the LED string, ask for the schematic up front — it’s cheaper as part of the original pass than as a separate job later.
When there’s an MCU in the loop
Plain analog drivers have no code. Dimmable and networked drivers often do — an 8-bit part handling PWM and a control bus, or a small ARM handling DALI. If that firmware is lost and you own the design, code readout may be possible depending on the part and how its protection is configured. STM8 and STM32 behave very differently here, and a part left at a low read-protection setting is a completely different job from one at the highest level. We’ll tell you which situation you’re in before quoting rather than after.
This service is for engineering work on designs you have the right to work on — recovering your own lost files, keeping installed fixtures serviceable, dealing with a discontinued controller, or analyzing a field failure. You confirm you hold those rights; we run every project under NDA and return or destroy samples on request.
What to send us for a quote
You don’t need to ship anything to get a feasibility answer. Send this and we can usually respond with scope, price range, and lead time:
- Clear photos of both sides of the board, whole board in frame, on a dark background, taken square-on rather than at an angle
- Close-ups of the controller IC marking, the transformer or inductor label, and any MCU
- Board dimensions and thickness, and the layer count if you know it
- Input and output ratings from the label or housing — mains or DC input, output volts and milliamps, dimming type
- One line on what you need: bare boards only, boards plus assembly, schematic, firmware, or everything
If the board is potted, say so and photograph the potting. It changes the answer more than anything else on this list. Photos with flash glare across the silkscreen are the most common reason we have to ask for a resend.
Where LED driver copies get hard, and where we say no
Full potting. A driver encapsulated in hard epoxy right to the connector can sometimes be opened without wrecking the traces and sometimes cannot. It depends on the compound and how much clearance sits between the copper and the fill. If you have two units, send both — one to sacrifice, one to keep as a reference.
House-marked primary controllers. We identify most of them from pinout and topology. A few we can only describe functionally, in which case we’ll propose a current equivalent and flag that the copy uses a substitute rather than the identical part. That’s a substitution decision you should be aware of, not something to quietly bury in the BOM.
Heavily damaged boards. A driver that failed by blowing the MOSFET and vaporizing a trace loses the very information we need. A charred board plus a good board is workable. A charred board alone means some nets get reconstructed from circuit reasoning instead of measurement, and we’ll tell you which ones.
Thermal design. A copy that duplicates the copper duplicates the thermal path. If the original ran hot in service and you want it to stop running hot, that’s a redesign, not a copy — the same distinction that comes up on boards used in sealed marine enclosures where ambient temperature is the real constraint.
Safety certification. We can reproduce the layout and the clearances. We can’t transfer the original certification. A copied mains driver going into a certified product needs to be re-evaluated by whoever holds that certification.
What drives the price
No two driver boards land in the same bracket. The variables that actually move the number: layer count, board area, component count, whether the board is coated or potted, how many custom magnetics need characterizing, whether you need a schematic or only fab data, whether firmware is involved, and whether you want assembled samples tested on a load. A small two-layer 20W driver with a marked controller sits at the low end. A 200W multi-channel potted driver with a custom transformer, a dimming MCU, and a schematic requirement sits well above it.
Turnaround follows the same pattern. Simple two-layer boards move quickly. Potting removal, magnetics reverse engineering, and firmware work each add real time, and we’d rather quote a longer honest lead time than miss a short one. Boards that share techniques with precision instrumentation work — tight tolerance measurement, careful net verification — take the time they take.
FAQ
Can you copy an LED driver PCB if the board is dead
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
Related reading
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