A microsection is the only way to see what actually happened inside a plated through hole, and it is the single most useful quality document a fabricator can give a customer. It is also easy to do badly: a section cut in the wrong plane, polished too far, or photographed at the wrong magnification can show a perfect hole that does not exist.
This is the checklist we use when reviewing a microsection report, either from our own shop or from a supplier being qualified.
1. When a microsection is required
First article of any new build. Any change of laminate, layer count, hole size, plating line or press cycle. Every Class 3 lot, at the sampling the drawing specifies. Any lot that fails electrical test in a way that suggests a barrel issue. Any field return where a via or hole is suspected.
For Class 2 production, a periodic section per lot or per shift is normal practice; the drawing or the quality agreement should state the frequency so it is not left to habit.
2. Sample preparation, step by step
Location. Section through the centre of the hole, in a plane that passes through the hole axis and at least two inner-layer connections. A section that misses the inner layers proves almost nothing.
Potting. Vacuum-impregnated potting resin, cured fully. Air bubbles at the hole wall are an artefact that looks like a void in the plating.
Grinding and polishing. Progressive grits, with the final polish fine enough that the copper-dielectric boundary is sharp. Over-polishing rounds the hole wall and makes the plating look thicker than it is at the knee.
Etching. A short etch to bring out the grain structure and the intermetallic, long enough to see the boundary, short enough not to attack the copper.
Magnification. Measure at 200x to 500x for plating and ring; use 1000x for intermetallic and micro-voids. State the magnification on every photograph, because a measurement without it cannot be checked.
3. What to measure
| Feature | Where measured | What goes wrong |
|---|---|---|
| Copper plating thickness | Mid-barrel and at the knee, at least two holes | Measured only at the surface, where it is thickest |
| Annular ring | Worst side of the pad, finished hole | Measured on the top view photo instead of the section |
| Etchback / pullback | Inner-layer connection, both sides | Not measured at all, or measured on one side only |
| Wicking | Along the inner-layer copper | Confused with etchback |
| Dielectric thickness | Between layer pairs | Measured at a via instead of in a clear field |
| Resin recession | At the barrel-to-dielectric junction | Ignored because the plating looks continuous |
| Voids and cracks | Whole barrel, especially the knee | Only the mid-barrel is photographed |
4. Acceptance, by class
Plating thickness, annular ring, etchback and voids all have separate limits for Class 2 and Class 3. The limits live in IPC-6012 and in the customer's drawing, and the drawing wins where it is stricter.
Two rules matter more than the numbers. First, measure at the worst point, not the average point. Second, evaluate after thermal stress: the standard requires the sample to survive thermal stress before sectioning, because the cracks that matter appear only after the barrel has been stressed.
5. The defects that show up most often
Barrel cracks at the knee. Usually a plating ductility or thermal expansion issue. On PTFE and high-Tg materials the Z-axis CTE is the driver; on FR-4 it is often plating thickness at the knee.
Voids in the plating. Contamination or poor hole-wall preparation. Small voids along the barrel can connect under thermal cycling and become an open.
Resin smear. Drilling too hot or desmear inadequate. It looks like a dark line between the copper and the inner layer, and it is a reject because the connection is mechanical rather than metallurgical.
Resin recession. The dielectric pulls back from the barrel, leaving a gap that plating bridges. It passes a pull test and fails thermal cycling.
Nailheading. The inner-layer copper is distorted by the drill. It reduces the effective ring and is a Class 3 concern even when the measured ring looks acceptable.
Glass protrusion. Glass bundles push into the hole and are not covered by copper. Common on some laminates if the drill parameters are wrong.
6. What a good report looks like
Photographs at stated magnification, one per measured feature, with a scale bar. Measurements tabulated per hole and per layer, with the acceptance limit printed next to each measured value. The panel and lot identification on every page. The thermal stress condition stated. A conclusion that references the drawing and the standard revision.
If the report shows one photograph of one hole and a sentence saying "plating thickness 25 µm, pass", it is a claim, not evidence.
7. Reading someone else's report
Check the section plane first: does it pass through the inner layers? Then check where the thickness was measured: knee or surface? Then check whether thermal stress was applied before sectioning. Then check whether the ring was measured on the section or on a top view.
Those four checks catch most of the reports that would not survive an audit. They take about two minutes.
8. Sampling that actually finds problems
One hole per panel is not sampling, it is ceremony. A workable scheme for Class 3 is: section holes from different board positions, including the panel edge and the centre, at least one via and one component hole, and repeat on the first three lots. After that, reduce to the drawing's requirement.
The failures that hurt are rarely uniform across a panel. They cluster at the edges, at the highest aspect ratio holes, and on the densest areas, and a single centre-panel hole will miss all three.
9. Checklist summary
Section plane through the hole axis and inner layers. Potting free of bubbles. Polish fine enough for a sharp boundary. Etch applied. Magnification stated. Plating measured at knee and mid-barrel. Ring measured on the section, worst side. Etchback, wicking, recession measured. Voids and cracks checked along the whole barrel. Thermal stress applied before sectioning. Panel and lot identified. Limits quoted from the current standard revision.
Eleven lines. A report that covers all eleven can be trusted; one that covers four is a picture of a hole.
Related reading
- Via Reliability Design Rules: Aspect Ratio, Ring and Protection
- IPC-A-610 Class 3: What Inspectors Actually Accept
- IPC-TM-650 Test Methods: Which Test Proves What
- IPC-6012 vs MIL-PRF-31032: Which Specification Applies
- J-STD-001 vs IPC-A-610: Process Standard or Product Standard?
- High reliability PCB manufacturing
- PCBA assembly services
