Failure Analysis of Field Returns
Why Field Returns Matter
A field return is the most informative object in a manufacturing organisation, because it is a real failure with a real history rather than a hypothesis about a process. Its value depends on the discipline applied to it: a return that is repaired and returned to the customer without an analysis teaches nothing, while one that is examined systematically can identify a design or a process problem that affects thousands of units. The analysis is therefore a routine with a defined sequence, not a task for the moment when someone has time.
Preserving the Evidence
The analysis begins before the unit is touched. The unit’s identity, its build date, its process data and its test results should be retrieved from the records, and the conditions of its failure should be documented: what it was doing, how long it had been in service, what the environment was, and how many units show the same behaviour. The unit itself should be photographed before any disassembly, and the board’s condition, including any contamination, corrosion or damage, should be recorded. A unit that has been handled, cleaned or partly disassembled loses evidence that cannot be recovered.
The Sequence of Tests
The sequence goes from the non destructive to the destructive. A visual inspection at magnification is first, followed by an electrical measurement of the supply and the key nets, which may localise the failure to an area. A thermal image under power shows a hot spot or an unexpected absence of one. X-ray shows the joints that cannot be seen, and a comparison with a known good unit of the same build is often the most informative single step. Only when the location is known should the destructive steps, such as a decapsulation or a section, be applied.
Looking for the Mechanism
The failure’s location is not the same as its cause. A cracked joint may be caused by a thermal expansion mismatch, an underfilled package may have failed for a mechanical reason, and a burnt component may be the victim of another component’s failure. The analysis should ask what the mechanism is, and it should test the hypothesis against the evidence: the crack’s path in a section, the contamination’s composition by analysis, the temperature that the part reached. The mechanism is what identifies the change that will prevent the recurrence, while the failed part is only the place where the symptom appeared.
Comparing Against Good Units
The comparison is the most practical tool. A returned unit compared with a known good unit of the same build, and both compared with a unit of a different build, shows what is different. The differences may be in the solder joints’ appearance, in the paste’s residue, in the mask’s condition, in the component’s date code or in the process data. The comparison should be systematic, and the units compared should be from the same or a stated different lot, because a difference between lots is itself a clue. The comparison often identifies the cause before any destructive test is needed.
The Interaction with the Process Data
The failure’s history is in the process records. The unit’s serial number leads to the board lot, the component lots, the machine and program, the profile and the test result. Where the analysis identifies a suspect process step, the records show whether that step was different for the affected units. Where the failure affects several units, the common factor in the records is the fastest route to the cause. The link between the analysis and the traceability system is what turns a single failure into a containment and a correction.
Recording the Result
The analysis’s output should be a record that identifies the failure’s mechanism, the evidence, the units affected and the action taken. Where the mechanism is not determined, that should be recorded too, with the hypotheses and what would be needed to test them, because an inconclusive analysis that is documented saves the next investigation from starting again. The record should be linked to the corrective action and to the containment, so that the history of the problem is available when a similar failure appears. A failure analysis that ends in a conclusion without a record leaves the organisation where it started.
Containment Before the Correction
An analysis that identifies a mechanism creates an obligation, which is to establish the extent of the problem before the correction is complete. The containment uses the traceability to define which units contain the suspect material or the suspect process, and it decides whether those units can be shipped, must be inspected or must be recalled. The containment has to be fast, because it is measured against the risk of another failure reaching the customer, and it is the step that most organisations find hardest under pressure. The traceability that supports it is the same traceability that a quality system requires for other reasons, which is one of the reasons the records are worth keeping.
FAQ
Why analyse a field return? Because it is a real failure with a real history, and it can identify a problem affecting many units that a repair would hide.
What should be done first? Retrieve the unit’s records and photograph it before any disassembly, since handling destroys evidence.
What is the most useful single step? Often a comparison with a known good unit of the same build, which shows what is different.
Why is the failure’s location not the cause? A burnt component or a cracked joint may be the victim of another mechanism, which is what the analysis has to identify.
What if the cause is not found? Record the analysis, the hypotheses and the tests needed, so the next investigation starts from where this one stopped.
Conclusion
Failure analysis is a disciplined sequence that preserves the evidence, compares against good units and identifies a mechanism rather than a part. Record the result. Analysis practice belongs to quality management, the manufacturing history it uses is described in PCB manufacturing, and the electrical measurements are part of PCBA testing. Field failures for a new product are investigated during prototype PCB assembly in 2026.
