Some board failures are not defects that escaped inspection. They are structures that grew after the board left the factory. Given moisture, ionic contamination and a sustained voltage difference, metal will migrate and form conductive paths between conductors that were correctly separated on the day of manufacture. The board that passed final test genuinely passed it.
Two related mechanisms
Electrochemical migration in its surface form produces dendrites: branching metal structures that grow across an insulating surface from cathode toward anode, visible under low magnification as fern-like or frost-like deposits. They can form in days under the right combination of humidity, contamination and bias.
Conductive anodic filament formation is the internal analogue. Copper migrates along a degraded glass-fibre-to-resin interface inside the laminate, typically between plated through-holes, producing a short buried where no external inspection can see it. CAF is slower, more insidious, and driven by the same three ingredients.
Contamination is usually the enabler
Humidity alone is rarely sufficient. What makes migration possible is ionic residue: flux activators not fully removed, chloride or sulphate from handling or the environment, plating salts, or residues from a cleaning process that was itself incomplete. Those ions provide the electrolyte that transport requires.
This makes cleanliness a measurable property rather than a matter of appearance. Ionic contamination testing and surface insulation resistance testing exist precisely because a board can look immaculate and be electrically dirty. Where a process changed — a new flux, a no-clean process treated as if it needed no control, a cleaning step removed to save cost — the change record is often where the answer sits.
Conformal coating and its gaps
Coating is the standard mitigation and it works when it is complete. It fails where it is thin, where it did not wet a surface, where it was masked and the mask was misplaced, and around tall components that shadow adjacent areas during application.
So the useful examination is not whether the board was coated but whether the coating covered the location that failed. Cross-sectioning through the failure site answers that directly, and coating thickness and adhesion at the failure are more informative than a general statement about the process.
Where the environment comes in
Condensation is the practical trigger in a great many service failures, and it is a thermal question rather than a humidity question alone. Equipment that cools below the dew point of the air around it wets internally, regardless of how dry the specification claimed the environment was. Enclosures that breathe, outdoor installations, refrigerated spaces and anything subject to a daily thermal cycle are all candidates.
Reconstructing that means reconstructing the installed conditions — enclosure sealing and drainage, heater or dehumidifier provision, ventilation, and the actual temperature and humidity history rather than the design assumption.
Distinguishing grown shorts from other failures
The distinction that matters is between a short that grew and one that was always there or was created by the event. Dendrites and filaments have distinctive morphology and composition: elemental analysis identifies the migrated metal, and its distribution shows direction of growth from cathode to anode.
Damage from an overstress event looks different — localised melting, vaporised conductor, damage centred on a junction rather than tracking between conductors. Where a board has burned, distinguishing the initiating short from the resulting fire damage is the same cause-versus-consequence discipline that governs electrical fire work generally.
Non-destructive first
The examination order matters because most of these techniques are destructive. Radiography and computed tomography locate internal features before anything is sectioned; CAF in particular is invisible externally and CT is often the only way to find it without guessing where to cut.
Once a site is located, cross-sectioning, scanning electron microscopy and elemental analysis characterise it. Cutting first and imaging afterwards frequently destroys the evidence, because the filament is a feature a few microns across in a specific plane.
Reproduction, carefully
Accelerated testing under temperature, humidity and bias can demonstrate that a given design and process combination is capable of the observed failure, which is a genuinely strong piece of evidence. It is also easy to overstate: showing that migration can be produced under aggressive acceleration does not establish that it occurred in service.
The persuasive version pairs reproduction with the field evidence — the same morphology, the same location, the same migrated species — and states the acceleration factors rather than presenting a chamber result as if it were a service observation.
Why it looks intermittent first
A grown short rarely arrives as a hard failure. A dendrite thin enough to bridge two conductors is also thin enough to fuse open when it carries current, which produces a fault that clears itself and then returns as growth resumes. Equipment behaves erratically for weeks, then works perfectly on a bench where the humidity is lower.
That signature is diagnostically useful rather than merely frustrating. A fault correlating with humidity, with time powered, or with a particular installed location — and disappearing under investigation — is characteristic of migration, and it is worth pursuing as such rather than being closed as unreproducible.
Where responsibility usually lands
These failures tend to resolve to process control and application rather than component quality. Design rules on conductor spacing and hole-to-hole distance, laminate selection, cleanliness control, coating coverage, and the suitability of an enclosure for its environment are each traceable to a decision and a document.
That is also why the manufacturing and process change records are worth requesting early. A migration failure across a population of boards almost always coincides with something that changed.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.