The definition of a double-sided PCB sounds trivial: copper on both faces, connected by plated through-holes. That sentence skips the single most important part of the process — the moment a hole is drilled, it connects nothing at all.
The drill passes through the dielectric of the laminate: epoxy resin and woven glass. Both are insulators. Drilling opens a channel through the board, but the wall of that channel is non-conductive. Only the severed copper foil edge at each hole opening is metal. To actually connect top to bottom, a continuous layer of copper has to be created from nothing on that insulating barrel.
That is what through-hole metallisation really does. It is not one operation; it is a chain with a strict causal order. This article follows that chain to answer three questions: which step decides whether it conducts, which step decides how long it keeps conducting, and why barrel copper is the acceptance item most often left out of a fabrication package.
(For what to inspect after drilling, see PCB drilling hole quality inspection. This article starts from "the hole is drilled and has passed inspection".)
Step one: desmear, because resin gets painted onto the barrel
Drilling is high-speed cutting and it generates heat. That heat softens the epoxy at the hole wall, and the drill flutes smear the softened resin along the barrel — particularly across the glass bundle cross-sections and any exposed copper. This redeposited film is called smear.
Why is it fatal? Because every subsequent metallisation step happens on the barrel surface. If the barrel is covered with a resin film, the copper you deposit grows on that film, not on the substrate. The film becomes a built-in separation plane, and the copper will delaminate along it under assembly reflow or thermal cycling in the field.
Desmear does two things, and the second is routinely overlooked:
- It removes the smeared resin residue so the true barrel surface is exposed.
- It micro-etches roughness into the resin. Once the resin is controllably attacked, the barrel gains a microscopically irregular topography, and the copper that follows relies on that topography for mechanical anchoring.
So desmear is not just cleaning — it is adhesion preparation. When it is inadequate, barrel copper can meet every thickness requirement and still fail bond strength. And that particular failure mode is invisible to thickness measurement: a microsection reads perfectly in-spec, and only thermal stress testing exposes it.
Step two: electroless copper, the step that decides whether it conducts
The barrel is now clean, roughened, and still insulating. Electroplating requires the workpiece itself to carry current, so plating is not yet possible. A first metal layer has to be formed without external current.
That is electroless copper. A chemical reduction reaction deposits an extremely thin but continuous copper film onto the activated barrel.
Two properties of this layer matter:
- It is very thin — far too thin to carry any current load or mechanical stress.
- It must be 100% continuous. A single uncovered stretch of barrel is an open circuit, and subsequent electroplating cannot grow across an insulating gap. The void propagates all the way to the finished board.
So judging whether a double-sided board "connects" is, mechanically, a judgement about whether electroless copper achieved continuous coverage. The classic way the industry examines that continuity is the backlight test: a microsection is ground to the hole centre, lit from one side, and observed from the other to see how much light passes through the barrel. More continuous copper means less light transmission. IPC-A-600 defines the observation as a graded criterion for electroless coverage quality — note this is an industry inspection method, and the applicable grade is set by the project acceptance documents.
This step decides conduction and decides nothing about reliability. A continuous but thin copper film can pass an electrical test once and still not survive a handful of reflow cycles.
Step three: electroplating, the step that decides how long it keeps conducting
The barrel now conducts, so it can be electroplated. The panel acts as the cathode in a copper plating bath; copper ions deposit onto the existing copper surface and build barrel copper from a film into a structure.
One geometric quantity governs this step — universally stated in design documents, routinely skipped during troubleshooting: aspect ratio, board thickness divided by hole diameter.
Why does it matter? Plating relies on convection and diffusion to transport copper ions into the hole. Near the opening, fresh solution is close and the field is stronger, so copper grows quickly. At mid-barrel — the deepest point, the hardest to exchange — it grows slowly. The higher the aspect ratio, the wider that gap becomes, and the microsection shows thinner copper at mid-barrel. Mid-barrel is precisely where thermal expansion stress concentrates.
Map this onto SKPCB's published double-sided specification and it becomes concrete. The double-sided PCB page lists a board thickness range of 0.4mm-3.2mm and a minimum hole size of 0.2mm. These two extremes are not freely combinable: 3.2mm thickness with a 0.2mm hole is an aspect ratio of 16, while 0.8mm with the same hole is 4. Two individually "in-spec" numbers from the same table can combine into plating tasks of entirely different difficulty. That is the engineering meaning behind the product page statement that hole diameter together with board thickness determines the aspect ratio, and with it the uniformity of copper deposition inside the barrel — and it is why hole size, board thickness and copper weight are reviewed together rather than one at a time.
Why the order must be "plate the hole first, image second"
The double-sided process chain contains one ordering rule that sets it fundamentally apart from single-sided fabrication: after drilling comes through-hole metallisation, and only then pattern transfer and etching.
Why can't that be reversed? Because electroless copper and electroplating are panel-wide. Every copper surface the solution touches gets plated; there is no way to plate only the holes. Etch the circuitry first and the plating step would thicken every finished trace along with the barrel, shifting trace widths and destroying the imaging accuracy you just achieved.
The correct order is therefore: drill → desmear → electroless copper → electroplate (panel copper and barrel thicken together) → pattern transfer → etch (both sides simultaneously) → solder mask, legend, surface finish → electrical test → profiling.
This order carries a consequence: because both sides are etched at once, the registration of top and bottom patterns relative to the drilled holes drives yield directly. Each side can be accurate on its own yet be offset relative to the other, which thins the annular ring on one side or breaks out of it entirely. Registration and profile tolerances are PCB-side parameters — see manufacturing tolerances.
Barrel copper thickness: the acceptance item most often missed
Now the point this article exists to make.
Nearly every fabrication package states copper weight clearly — "1oz copper" or "2oz copper". That number almost always refers to surface copper: foil weight plus panel plating. Barrel copper thickness is a separate acceptance object, governed by different mechanisms and measured differently:
| Object | Governed by | How it is verified |
|---|---|---|
| Surface copper thickness | Base foil weight + panel plating | Surface gauging, microsection |
| Barrel copper thickness | Electroless continuity + in-hole plating uniformity (aspect-ratio dependent) | Microsection only, taken at mid-barrel |
Why doesn't in-spec surface copper imply in-spec barrel copper? Because ion exchange inside the hole is worse than at the panel surface, and the deeper and narrower the hole, the wider the gap. Surface copper can reach target while mid-barrel copper on the same panel, in the same plating cycle, stays thin.
This is why IPC-6012 specifies minimum barrel copper thickness by acceptance class, with tighter requirements at higher classes. SKPCB's quality system accepts work to IPC-6012 Class II and Class III — the specific minimum barrel copper value depends on the standard revision applicable to your project and is confirmed in engineering review. This article deliberately does not quote a representative figure, because the value moves with revision and class, and a hard-coded number would mislead the judgement.
Three practical recommendations for the design side:
- State the acceptance class in the documentation. "Per IPC" is not a specification; Class II and Class III impose different barrel requirements.
- Declare barrel copper separately. Writing "1oz copper" does not cover it.
- Give the board-thickness and minimum-hole-size combination, not two independent limits. Aspect ratio is a quotient; only the combination defines a reviewable plating task.
(For surface copper options and design rules see PCB copper thickness options and design rules; for plating inspection and release criteria see PCB electroplating quality inspection and release. Laser-drilled microvias follow a different aspect-ratio regime — see laser-drilled microvia plating and aspect ratio. This article covers mechanically drilled through-holes in double-sided boards only.)
The causal chain in one block
- Desmear decides whether copper adheres (get it wrong and in-spec thickness still delaminates).
- Electroless copper decides whether it conducts (a discontinuity is an open circuit and cannot be recovered downstream).
- Electroplating decides how long it conducts (thickness and mid-barrel uniformity drive thermal cycling life).
- Aspect ratio is the geometric constraint running through both plating steps, set by the combination of board thickness and hole diameter, never by either value alone.
For a manufacturability review against a specific thickness, hole size and acceptance class, submit Gerber/ODB++, drill and profile data, board thickness, copper weight, surface finish and acceptance class together — see capabilities or get a quote. Assembly-side interface requirements are on the PCBA services page.
FAQ
Q: If the barrel isn't conductive, why not just electroplate it directly?
Electroplating requires the workpiece to conduct current so copper ions can deposit on it. A freshly drilled barrel is resin and glass — an insulator — so no current flows and nothing plates. That is why electroless copper comes first: a chemical reduction reaction that needs no external current deposits a thin continuous copper film, turning the insulating wall into a conductive one. It also explains why electroless continuity is the precondition for the whole chain: wherever it is discontinuous, electroplating cannot grow either.
Q: Is desmear just a cleaning step? Can it be shortened?
No. Desmear does two jobs: it removes the resin residue that drilling heat smeared onto the barrel, and it micro-etches microscopic roughness into the barrel. The second job is where the copper layer's mechanical anchoring comes from. When desmear is skipped or inadequate, copper may grow on a residual resin film, creating a separation plane. The characteristic of this defect is that thickness measurement cannot detect it — a microsection reads fully in-spec while bond strength is inadequate, and only thermal stress testing exposes it.
Q: Is barrel copper thickness the same as the "1oz copper" I specified?
No. Stated copper weight normally means surface copper: foil weight plus panel plating. Barrel copper thickness is a separate acceptance object, determined by electroless continuity and in-hole plating uniformity, the latter driven by aspect ratio (thickness ÷ hole diameter), and measurable only by microsection at mid-barrel. On the same panel in the same plating cycle, surface copper can be on target while mid-barrel copper is thin. Barrel copper should therefore be declared separately in the documentation, with the IPC-6012 acceptance class named explicitly.
Q: Why must a double-sided board be plated before imaging?
Because electroless copper and electroplating are panel-wide: every copper surface the solution reaches is thickened, and there is no way to plate holes without plating the surface. If the circuitry were etched first, plating would thicken every trace, shifting widths away from design intent and destroying imaging accuracy. The order is therefore fixed: drill → desmear → electroless copper → electroplate → pattern transfer → simultaneous two-side etch. That order also imposes a requirement of its own: the registration of both patterns relative to the drilled holes directly affects annular ring integrity.
Q: Can I use 3.2mm board thickness with a 0.2mm hole?
The double-sided PCB page publishes a thickness range of 0.4mm-3.2mm and a minimum hole size of 0.2mm, but those two extremes do not form an automatically valid combination. 3.2mm with 0.2mm gives an aspect ratio of 16, and mid-barrel plating uniformity is far more demanding than for a thin board with the same hole. Actual feasibility depends on copper weight requirements, acceptance class, and hole count and distribution, confirmed in engineering review against project data. Both values sitting inside the published range is not sufficient on its own.