PCB Delamination

PCB delamination is the separation of the bonded layers inside a printed circuit board, either between copper foil and laminate or between the epoxy resin and the woven glass fiber that reinforces it. It can appear as visible blisters, whitish "measling" spots, or raised areas on the board surface, or it can remain completely hidden as an internal debonded interface between resin and glass. A circuit board is a laminated composite, and like any laminate, plywood, skis, aircraft panels, its strength and its insulation both depend on the plies staying bonded. Once layers separate, the board's mechanical integrity and its electrical insulation are compromised at the same time.

Causes

The dominant trigger is trapped moisture flashing to steam during a hot process step. Laminate resins absorb ambient humidity slowly over weeks; when a stored board then hits reflow or rework temperatures, that absorbed water vaporizes and expands violently, prying the layers apart from inside, the so-called popcorn effect, named for the audible crack it can produce. The second trigger is simply exceeding the laminate's glass-transition temperature (Tg) for too long, softening the resin until bonds fail. The third is cumulative: repeated thermal cycling stresses the resin-glass bond through the mismatch in expansion rates between copper, resin, and glass, and years of heat-up/cool-down cycles can propagate a small void into a spreading separation. On the shop floor, hand soldering with excessive heat dwelling on one pad, and aggressive hot-air rework around large components, are the classic self-inflicted causes.

Why it is dangerous

Beyond mechanical weakness, a debonded glass-resin interface opens a highway for conductive anodic filament (CAF) growth: under voltage bias and humidity, copper salts migrate along the separated glass bundle and gradually form an internal conductive path between vias or traces that were never meant to meet. The resulting short is buried inside the board, invisible to inspection, often intermittent and humidity-dependent, and effectively unrepairable, the failure mode shares its electrochemical family with surface dendrite growth, but happens where no cleaning can reach. Delamination also degrades impedance control and can crack internal via connections as the separated region flexes.

Prevention and relevance to hashboard repair

Prevention is mostly discipline. Bake boards that have been stored unsealed before any reflow or major rework (typical practice is hours at around 105-125 °C, per the laminate and moisture-sensitivity guidance), control soldering temperature and dwell time, use preheaters so rework heat does not create steep local gradients, and specify high-Tg laminate for assemblies that live hot. All of this lands directly on mining hardware: a hashboard runs near the top of its comfortable temperature range for years, and repairing one means reworking large BGA-adjacent ASIC packages with serious heat. An amateur rework with a bare hot-air gun, no preheat, no bake, no thermal profile, is one of the most common ways a repairable board becomes scrap: the chip swap succeeds and the board dies weeks later from the delamination the rework caused. This is why proper repair benches use controlled preheating and profiled heating, and why a board with suspected internal damage deserves professional triage through a service like D-Central's repair intake rather than a third round of improvised heat. See also thermal fatigue for the sibling failure mode that attacks the solder joints instead of the laminate.

Detecting internal delamination is its own discipline. Surface blisters announce themselves, but hidden separations require indirect methods: scanning acoustic microscopy images debonded interfaces non-destructively, cross-sectioning sacrifices a board to confirm a suspected failure mode, and in the repair shop the humble signs, a board that flexes with a faint crackle, a whitish halo around a reworked area, a fault that follows humidity, are often the only affordable evidence. When several boards from one batch show the same symptoms after the same rework process, the process is the suspect, not the boards.