If the switching power supply in a machine you support has died and the OEM stopped building it years ago, a power supply PCB copy is usually the fastest path back into production. We take your physical board — working, dead, or partially populated — and rebuild a manufacturable file set from it: copper on every layer, drill data, a BOM with real part numbers, and the transformer and inductor specs. You get files you can send to a fab and an assembler.
Power boards are their own animal, so this page sticks to them. Signal boards are mostly about connectivity. A power board is about connectivity plus copper cross-section, clearance, thermal paths, and magnetics you can’t read off a marking.
What makes a power supply PCB copy different from a logic board
On a controller board, if you recover every net correctly, you’re basically done. On an SMPS, a net-perfect copy can still fail, and here’s what an engineer is actually watching for while the job runs.
Copper weight and trace cross-section
Power boards frequently run 2 oz, 3 oz, or heavier copper, sometimes only on the outer layers with 1 oz inside. We measure the finished copper thickness rather than assuming it, because a 3 oz trace rebuilt in 1 oz is a slow-motion failure — the trace works on the bench and cooks in a hot cabinet. IPC-2152 is the reference for how much current a given cross-section actually carries at a given temperature rise, and it’s the sanity check we apply to the wide traces before releasing files.
Creepage, clearance, and the primary-secondary barrier
The isolation gap between primary and secondary is a safety dimension, not a layout preference. We measure it — across the board surface, around the optocoupler and the Y-cap, and under the transformer footprint — and hold it in the rebuilt layout. Any slot or routed cutout in the barrier gets carried over exactly. If someone later asks us to shrink the board, that gap is the one thing that doesn’t move.
The magnetics nobody documented
The flyback transformer is the part that kills most naive copies. It usually has a house part number that means nothing outside the original factory. To reproduce it we need turns counts, winding order, wire gauge, interleaving, tape layers, core material and gap. That means measuring inductance and leakage inductance first, then, on one sacrificial unit, unwinding it and documenting the build sheet layer by layer. There is no shortcut, and no photo will give you this.
Thermal reality
We record where the heatsinks bolt down, which pads are thermal relief and which are solid, how big the copper pours around the MOSFET and the diode are, and whether there are unplated thermal windows. Copper pour on a power board is a heatsink. Redraw it smaller and the part runs hotter.
How we run the job, step by step
- Intake and documentation. High-resolution scans of both sides fully populated, component locations logged, and any silkscreen or date codes recorded before anything is touched.
- Electrical measurements while it’s intact. Transformer and inductor inductances, resistor values in-circuit where possible, optocoupler and controller pinout confirmation.
- Desolder and identify. Every part comes off. Electrolytics get read for capacitance, voltage, ripple rating, and case size. Marked-over or house-numbered parts go through component identification using package, function in-circuit, and behavior.
- Bare board imaging and layer separation. Solder mask off, scan, then grind or delaminate down layer by layer for multilayer boards, aligning each image to a common set of tooling features. Two-layer boards skip most of this; four- and six-layer power boards don’t.
- Copper measurement. Finished copper thickness, minimum trace and gap, drill diameters, annular rings, and the layer stackup including dielectric thicknesses.
- Rebuild in CAD. Copper is drawn as copper, not traced as a bitmap — every polygon, pour, and via lands on real coordinates so the file survives editing later.
- Netlist verification. The rebuilt netlist gets compared point-to-point against the original bare board. Discrepancies get resolved on the board, not in the software.
- Fab, build, and bench test. We build a prototype and put it on a load. Startup, regulation, ripple, cross-regulation on multi-output units, and thermal soak at full load — measured against the original where a working original exists.
That last step is the one that separates a file set from a working power supply. Bench data on a resistive load and a scope trace of the switching node is worth more than any accuracy claim.
Which power supply boards are harder, and which we sometimes decline
Being straight about this saves everyone a week.
| Board type | Difficulty | Why |
|---|---|---|
| Two-layer linear or simple flyback | Straightforward | Few layers, generic parts, magnetics often off-the-shelf |
| Four-layer SMPS, 2 oz copper, discrete controller | Moderate | Inner layers need separation; copper weight and creepage must be measured |
| Multi-output server or telecom supply | Higher | Multiple magnetics, tight cross-regulation, digital control loop |
| Digitally controlled PFC / LLC with a locked MCU or DSP | Highest | Hardware copies fine; the board is dead without the firmware |
| Fully potted or epoxy-encapsulated module | Case by case | Removing potting can destroy the copper; sometimes only one unit survives, sometimes none |
| Custom planar transformer in the PCB stackup | Case by case | Windings are the layers; recoverable, but the tolerance stack is unforgiving |
Two specific cases change the answer more than anything else. First, potting compound: silicone we can usually work through, hard-filled epoxy sometimes takes the pads with it, and we’ll tell you the odds before you ship rather than after. Second, a digital controller with read protection enabled. The board copy is unaffected, but the unit won’t regulate without the code inside that part. That becomes a separate firmware question, and feasibility depends entirely on the specific device and protection level.
We also won’t quote a copy of a board that’s obviously a current, actively sold commercial product for someone with no claim to it. You need to hold the rights to the design or a legitimate maintenance interest in the equipment, and every project runs under NDA. That’s the whole of it.
What to send us, and what you get back
To quote a power supply PCB copy we need very little, but it has to be the right little:
- Clear photos of both sides, whole board in frame, on a dark non-reflective background, in even light. No flash glare on the mask.
- Board dimensions — length, width, and overall thickness — and layer count if you know it.
- The markings on the main controller IC and the transformer, close up and legible. These two parts drive most of the risk.
- Input and output ratings: AC or DC input range, each output rail’s voltage and current.
- One line on what you need — files only, files plus a working prototype, or a full production run.
Back comes a feasibility read, a lead time, and a price. Then, on the job itself: Gerbers with the full stackup and copper weights called out, drill files, an editable CAD project, a schematic if you ordered one, a BOM with manufacturer part numbers and noted substitutions, a pick-and-place file, and the transformer build sheet. The bundle is the same idea as our general PCB copy service deliverables, with the magnetics documentation and thermal notes added on top.
What drives the price on a power board
Nobody can quote this from a keyword. The real drivers are layer count, board area, distinct component count, how many custom magnetics need characterizing, whether the board is potted, whether firmware is involved, and how much of the deliverable set you want. A single-output two-layer supply with a catalog transformer sits at one end. A four-layer PFC front end with two custom magnetics and a locked controller sits far from it.
Obsolescence is the other cost you should be weighing. If the original PWM controller is end-of-life, an exact copy just buys you time until the last broker stock runs out. In that situation we’ll flag it and suggest either a stock buy on the copy as-is, or replacing the obsolete components with current equivalents and re-verifying the loop — usually cheaper than a ground-up redesign, and the board stays interchangeable in the chassis.
FAQ
Can you copy a power supply board if it’s completely dead?
Yes, in most cases. Copper topology, layer stackup, and component identity survive electrical failure. What we lose is baseline reference data — we can’t measure the original’s ripple or regulation to compare against. A shorted or exploded part also has to be identified from its footprint and circuit function instead of its marking, which takes longer. Two units, one dead and one working, is the ideal.
Do I have to send you the transformer too?
Yes, and ideally two. Custom magnetics carry house part numbers that no distributor can cross-reference, so the only way to reproduce one is to measure it and then unwind a sacrificial sample to document turns, wire gauge, winding order, and core gap. If you send only one unit, we’ll measure non-destructively first and discuss the tradeoff before opening it.
How long does a power supply PCB copy take?
A two-layer board with standard parts is typically a matter of days. A four-layer SMPS needs layer separation plus magnetics characterization, so meaningfully longer. Add time again for a built and load-tested prototype, and more if a locked controller or potting is in the picture. We give a real date after seeing photos, not before.
Will the copy pass safety certification?
The copy preserves the original’s creepage gaps, isolation barrier,
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