Area-array packages moved every solder joint underneath the component, where no visual inspection can reach. That was a deliberate trade for density and electrical performance, and it means the joints most likely to matter in a failure analysis are the ones nobody has ever looked at. Several of the characteristic defects also pass electrical test at room temperature and only open under thermal or mechanical load.
Head-in-pillow
The defect is exactly what the name describes: the ball and the paste deposit each melt and wet their own side but never coalesce into one joint. Under X-ray the ball looks present and roughly the right shape. Electrically it may conduct through incidental contact, so the board passes test and ships.
The causes are well characterised — package warpage during reflow separating the ball from the paste at the critical moment, oxidation preventing wetting, insufficient paste volume, or a reflow profile that mistimed the two. Warpage is the dominant one, and it is a function of package construction, board design and the thermal profile together rather than any single party's choice.
Voiding, and when it matters
Voids are gas porosity trapped in a joint as flux volatiles escape during solidification. Some voiding is normal and acceptable, which is why the guidance in the IPC documents for BGA assembly sets criteria rather than demanding none.
What matters is size, position and consistency. A void at the package or board interface, where stress concentrates, is more consequential than one in the bulk of the joint. Large voids reduce effective cross-section, raising local current density and thermal resistance. And a population of boards with abnormal voiding usually points at a process — paste, profile, stencil design or via-in-pad without adequate fill — rather than at one bad assembly.
What each technique can actually resolve
Two-dimensional X-ray is fast and shows ball presence, gross voiding, bridging and misregistration. It is a projection, so it struggles to say where in the joint's height a feature sits, and head-in-pillow frequently looks normal.
Computed tomography resolves that height information and can identify separation planes and interface voiding without cutting. Cross-sectioning remains the reference method for intermetallic structure, crack path and true void geometry, but it is destructive and only examines the joints the section happens to pass through — which is why the imaging comes first and selects the cut.
Dye and pry, used honestly
Dye penetration followed by forced separation is a cheap and genuinely useful technique: dye wicks into any pre-existing crack, and after separation the stained area maps the crack that existed before the test.
Its limitation is that it destroys the joint and cannot distinguish a crack it created from one it found unless the staining is unambiguous. Performed before any other examination, it forecloses cross-sectioning. Performed on a sample after imaging has established what to expect, it corroborates well.
Rework history
A reworked BGA is a different object from an as-built one. Rework subjects the package and the board to additional thermal cycles, frequently with less profile control than production reflow, and it may introduce a different alloy or flux chemistry. Pad damage from removal, residual intermetallic from the original joint, and mixed alloys are all common and all detectable.
So establishing whether a site was reworked is an early question, and it is usually answerable from the joint microstructure and from assembly records together. It also reallocates responsibility, often away from original manufacture.
Design contributions
Board-side choices influence the outcome as much as the assembly process. Pad geometry and whether the pad is solder-mask-defined or non-solder-mask-defined changes where stress concentrates and where cracks initiate. Via-in-pad without proper fill and cap provides an escape path that promotes voiding. Board thickness and local copper distribution drive warpage and thermal gradients.
These are recoverable from the design files, and comparing them against the package manufacturer's recommendations is straightforward. A deviation from those recommendations is a finding that sits with the designer rather than the assembler.
Testing the crack, not just imaging it
Imaging shows a crack; it does not show whether that crack explains the reported symptom. Bridging the two usually needs the joint exercised. Resistance measurement across the suspect net while the assembly is heated, cooled or flexed can reproduce an intermittent open and tie it to a specific location.
Boundary-scan and in-circuit techniques can localise an open to a net without any physical access, which is valuable on packages where probing is impossible. Doing this before destructive work means the cross-section is aimed at a joint already known to misbehave, rather than at the one that happened to look worst on the X-ray.
Preservation
The board intact, unpowered, and not reworked. Any attempt to repair the failure destroys the evidence of it, and the instinct to reflow a suspect BGA to see whether the fault clears is the single most effective way to lose the answer.
Where the assembly was in a system, the mechanical context is worth keeping too — mounting, standoffs, heatsink clamping and anything imposing load on the board, since those forces act directly on the joints under the package.
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.