Few things frustrate a warranty organisation more than a unit that fails in service, gets returned, passes every functional test on the bench, and ships back to the customer. Labelled "no fault found," it is quietly treated as a customer error or a software glitch. In electronics failure analysis, that classification is usually wrong, and the reason is structural rather than mysterious.

Why the bench cannot see it

A solder joint with a fatigue crack propagating through it does not behave like an open circuit. At room temperature, with no load and no vibration, the crack faces remain in contact and the joint conducts normally. Warm the assembly, flex the board, or subject it to vibration, and differential expansion pulls those faces apart. The circuit opens, the fault appears, and when the unit cools on a test bench, it closes again.

The failure is therefore conditional on an operating environment the bench does not reproduce. "No fault found" is very often an accurate statement about the test and an inaccurate statement about the hardware.

Why the joints crack in the first place

Solder is a soft metal operating at a high fraction of its melting temperature. It creeps under sustained load and accumulates fatigue damage under cyclic strain. Every power cycle and ambient swing strains the joint between a component and a board that expand at different rates. Strain concentrates at the outer corners of larger packages, which is where cracks typically initiate and why BGA corners are the first place an analyst looks.

The mechanism is well characterised in the reliability literature, including work on failure behaviour during thermal fatigue (IPFA 2009) and on the compounding effect of sequential thermal and vibration exposure (Springer, 2008), which is closer to what most field hardware actually experiences than either stress alone.

Process conditions established at assembly change the timeline. Voiding reduces the load-bearing cross-section and concentrates both stress and heat; excessive intermetallic growth creates a brittle layer that fractures under shock rather than fatiguing gracefully. Work on vacuum reflow and its effect on voiding and thermal fatigue performance illustrates how much of field reliability is decided at manufacture (SMTA proceedings).

What actually finds it

The investigation runs in a deliberate order. X-ray or CT inspection first, because it is non-destructive and can reveal cracks, voids, and head-in-pillow defects under packages that cannot be inspected visually. Then environmental testing — thermal cycling or vibration while monitoring electrically — to reproduce the fault under conditions matching service rather than the bench. Only then does the analysis become destructive: dye-and-pry to distinguish pre-existing cracks from damage created during removal, followed by metallographic cross-sectioning through the suspect joint to expose the crack path and microstructure.

That sequence answers the question that decides the business outcome: is this an outlier or a systemic defect? A handful of failed units establishes the mechanism. Comparison against unused units from the same lot, and against units from other lots and date codes, establishes scope. The difference between the two is the difference between a warranty reserve and a recall.

The practical instruction

Do not rework suspect assemblies. Reflowing, desoldering, or repairing a returned unit destroys the crack morphology that identifies the mechanism and forecloses the analysis entirely. Quarantine failed units with their production records, date codes, and traceability intact.

For the full mechanism set and examination protocol, see our specialization area on solder joint failure analysis.