With a metal core PCB copy, layer count changes the job more than board size does. A single-layer insulated metal substrate (IMS) board is mostly a dielectric measurement problem — one copper layer, one thin dielectric, one aluminum or copper base. A two-layer or multilayer metal core board adds vias that cannot pass through the metal, which means the copy work shifts toward figuring out how the original designer routed around a conductive base plate.
That distinction drives everything below: how the board gets sectioned, what data we can recover, and how honest we can be about matching thermal performance.
Why the metal base changes the copy process at all
On a normal FR-4 board, layer separation is a controlled etch-and-image cycle. You strip solder mask, image the outer copper, remove it, and work inward. A metal core board resists that flow because the base is aluminum (usually 1060 or 5052) or copper, and it doesn’t behave like glass-reinforced epoxy in the same chemistry.
So the sequence is different. We image the copper first, then mechanically separate the copper-and-dielectric sandwich from the metal base rather than trying to etch through it. The dielectric layer — typically a ceramic-filled polymer somewhere in the 75 µm to 150 µm range on lighting boards — is the part that actually matters for performance, and it’s also the part that gets destroyed if you’re careless.
Three things are recorded before any destructive step:
- Total board thickness and base metal thickness, measured with a micrometer at several points, because aluminum stock tolerance is loose and one reading isn’t enough.
- Copper weight on each layer, from cross-section under a microscope. Metal core boards commonly run 1 oz to 3 oz, and a 3 oz copy built in 1 oz will fail thermally even if the netlist is perfect.
- Dielectric thickness and appearance — color, filler visible under magnification, whether it looks like a standard thermally conductive prepreg or a higher-performance ceramic-loaded system.
We can measure dielectric thickness accurately. We cannot read the thermal conductivity number off a piece of cured dielectric. That gets estimated from the application, the copper weight, and the power dissipation implied by the components — and it gets stated as an assumption in the report, not as a fact.
Single-layer IMS boards: the fast case
The majority of metal core boards that come through the lab are single-layer: LED strips, downlight modules, high-bay engines, motor driver boards with a few TO-263 parts, and power supply daughtercards. One copper layer, no vias through the dielectric, all components on top.
These are the quickest boards we handle, sometimes faster than a two-layer FR-4 board of the same size, because there is no inner-layer registration problem. The copper pattern is visible after solder mask removal and can be scanned directly. Work concentrates on:
- Scanning both the copper pattern and the solder mask opening pattern separately, since thermal pad openings on IMS boards are frequently non-standard and matter for reflow.
- Recovering the board outline, mounting hole positions, and any milled or scored edges — metal core boards are usually routed, not V-cut, and the tooling matters.
- Identifying the LEDs or power devices. This is the step that most often stalls the job, not the copper.
The component identification problem on LED boards
An LED package marking tells you very little. Two visually identical 3030 packages can differ by bin, CCT, forward voltage, and flux. If the customer needs the copy to produce the same light output at the same drive current, we need either the original purchase records, a working sample to measure, or acceptance that the rebuild will be a functional equivalent rather than a photometric match. We say that up front rather than after delivery.
The same logic applies to power MOSFETs and diodes in DPAK and D2PAK on motor and driver boards — the topmark is often a house code. Cross-referencing gets us to a plausible part; electrical measurement on a live board gets us to a defensible one.
Two-layer and multilayer metal core: where the real work is
Once there are two or more copper layers, the designer had a routing problem you now have to reconstruct. The metal base is conductive, so a plated through-hole cannot simply pass through it. Real boards solve this in a few recognizable ways, and part of the job is determining which one you’re looking at.
| Construction | What you see | Copy difficulty |
|---|---|---|
| Single-layer IMS | One copper layer bonded to base, no vias in the metal | Low — dielectric and copper weight are the main unknowns |
| Double-sided FR-4 bonded to aluminum | A conventional 2-layer board thermally bonded to a plate, often with adhesive | Moderate — separates into a normal 2-layer copy plus a mechanical carrier |
| 2-layer true metal core | Copper on both sides of the metal, vias insulated from the base by a clearance ring in the metal | High — via-to-metal clearance and drill sequence must be measured, not guessed |
| Multilayer hybrid (4+ layers on a metal base) | FR-4 or ceramic-filled stackup laminated to a heavy base, sometimes with a thermal slug or coin | Highest — needs full layer separation plus X-ray, and the bonding step is nontrivial to reproduce |
For the last two rows we X-ray before touching anything. On a true 2-layer metal core board, the X-ray shows whether the pre-drilled and backfilled clearances in the base metal line up with the through-holes, and how much annular clearance the designer allowed. Get that dimension wrong on the rebuild and you have a board that passes continuity and fails hipot.
Multilayer hybrids get the same layer-by-layer treatment we’d apply to any multilayer board compared against a single-layer one: image, scan, verify, remove, repeat, with the netlist rebuilt from the layer set and cross-checked against a continuity map taken from the intact board before separation. The extra wrinkle is that the metal base often acts as a thermal path, sometimes as a shield, and occasionally as a ground return through a deliberate connection. Which of those it is has to be confirmed electrically, not assumed.
Thermal vias, coins, and slugs
Boards with a thermal coin or press-fit copper slug under a hot device need dimensional capture beyond the copper artwork — slug diameter, protrusion, and whether it’s a press-fit or plated feature. These details don’t appear in Gerber data. They go into a separate mechanical drawing in the deliverable set, which is why the file package for a metal core job is usually larger than for an equivalent FR-4 board.
What we need from you, and what comes back
To quote a metal core PCB copy accurately, send this much:
- Clear photos of both sides against a dark background, whole board in frame, plus one close-up of any device you care about matching.
- Board dimensions and total thickness, plus base metal thickness if you can measure it.
- The main chip or LED marking, typed out as text — photos of laser marks are often unreadable.
- One line on deliverables: Gerber only, Gerber plus BOM, full schematic, or a working rebuilt board.
- Whether the board must match thermal performance or only function.
You get back a feasibility note, a lead time, and a price. The feasibility note says specifically what we expect to recover and what will be an engineering estimate — dielectric material class is nearly always in the second category.
Deliverables typically include Gerber and drill data, a stackup drawing with measured copper weights and dielectric thickness, a BOM with confidence flags on the parts we couldn’t confirm from markings alone, and a mechanical drawing for the base metal, routing, and any slug features. Schematic and netlist come with the higher tiers. If firmware sits on the board — common on smart LED drivers and motor controllers — that’s a separate scope; the customer needs to hold the rights to the design, and we run these projects under NDA.
Honest limits on metal core work
A few cases are genuinely harder, and one or two we’ll turn down:
- Unknown high-performance dielectrics. We can measure thickness and dielectric strength behavior on a coupon. We can’t reverse the formulation. If your board only works because of a specific 3 W/m·K or better material, we’ll tell you the copy needs a spec’d substitute and validation testing.
- Boards already thermally damaged. Delaminated dielectric, warped base plate, or a scorched area near a failed device destroys exactly the region you most need to read. A second sample helps enormously.
- Potted or conformally coated assemblies. Removable, but slower, and the removal chemistry sometimes takes the solder mask legend with it, which affects silkscreen recovery.
- Single-sample multilayer hybrids. Layer separation is destructive. With one board and four layers on a metal base, there’s no second chance to re-image a layer you lost. We’ll say so before we start.
- Photometric matching without a reference. Covered above — worth repeating because it’s the most common source of disappointment on LED boards.
None of this is unique to metal core boards in principle, but the thermal coupling makes the consequences of a small error larger. A 10% dielectric thickness error on an FR-4 signal board is usually invisible. On a 30 W LED module it changes junction temperature and it changes lifetime.
How layer count affects cost and schedule
Price and lead time on metal core jobs track, roughly in order of impact: number of copper layers, whether the base has insulated via clearances, component count and package difficulty, board area, and whether you need a rebuilt working sample instead of files only. A single-layer LED module with twenty identical LEDs and one connector is a short job. A four-layer hybrid with a BGA controller, a thermal coin, and one available sample is a different conversation entirely.
Boards with tight impedance requirements on the signal layers add verification steps, similar to what an
Send the chip marking or two photos. You get feasibility, lead time and price within 24 hours, and the check costs nothing.Working on a board like this?
Related reading