QE Skill Enhancement (31) | Field Failure Analysis (FA): From Returned Units to Root Cause

1. A Case of "Dismantling the Failed Units"

A manufacturing company received 12 faulty units returned by a customer. To quickly respond to the customer, the after-sales team directly dismantled 3 of them for inspection and found a bulging capacitor on the motherboard. They concluded that it was a "capacitor quality issue" and demanded compensation from the supplier, along with a brand change. Three months later, a similar batch of failures occurred again, but this time the bulging position and the component manufacturer were different. The conclusion, however, remained the same: "supplier component defect."

Later, the Quality Engineer (QE) reorganized the analysis: the remaining 9 returned units were archived by serial number, and non-destructive inspections such as visual and X-ray were conducted first. It was discovered that 7 of the units had cracks at the same solder joint. Further analysis using dye penetrant and cross-sectioning confirmed that the issue was due to a cold solder joint caused by a low temperature in a specific zone of the reflow soldering profile. The cracks expanded under thermal cycling and eventually led to an open circuit. The true root cause was the process parameter, not the component.

The cost of the first failed analysis is very typical: evidence was not preserved before disassembly. Once destructive methods are employed, the phenomena cannot be reproduced, and the mechanism cannot be pinpointed, leaving only guesswork.

2. Key Principles: FA is an Evidence Chain, Not a Blame Game

（1）Three-Stage Localization, Order Cannot Be Reversed.

The goal of failure analysis is to connect the "phenomenon—mechanism—root cause" in three stages: the phenomenon is "not working," the mechanism is "a solder joint open circuit," and the root cause is "the measured temperature in the reflow soldering zone is below the process lower limit." Skipping the mechanism and directly attributing the root cause (e.g., "supplier component defect," "operator error") is an unfounded attribution in terms of evidence. If the corrective action is misdirected, similar failures will inevitably recur.

（2）Failure Physics Determines the Sequence of Analysis Methods.

Observation methods range from non-destructive to destructive: visual → X-ray/CT → electrical performance → decapsulation and delayering → cross-sectioning. Stop at any step that can identify the mechanism; do not escalate the method just to "see everything." Destructive methods should only be used when non-destructive methods fail to identify the mechanism, and at least 2 undamaged samples from the same batch must be retained for verification and comparison.

（3）The Judgment Chain Must Be Reversibly Verified.

For a failure mechanism to be valid, three conditions must be met simultaneously: ① the failure location matches the predicted mechanism location; ② the stress conditions of the failed units can reproduce the mechanism; ③ the characteristic does not appear in non-failed units. If any of these conditions are missing, the conclusion can only be downgraded to "suspected." This is also the interface between FA and FMEA—once the mechanism is confirmed, the detection (D) and control measures for the corresponding failure mode can be modified.

3. Practical Steps (5 Steps, Each with Criteria)

Step 1: Sample Collection and Evidence Preservation.

• Sample Quantity: For a single batch of returned units with the same phenomenon, retain ≥5 units for analysis (1-2 units for non-destructive localization, 2-3 units for destructive analysis, and ≥2 units left untouched); for batch failures (same batch ≥3 units with the same phenomenon), freeze the entire batch by batch number and do not use individual units.
• Time Limit: Complete registration, photography, and serial number sealing within 48 hours of sample collection; do not power on or disassemble the units before sealing.
• Preservation Requirements: Use anti-static packaging (wrist strap resistance 1 MΩ to 10 MΩ, contact surface static voltage <100 V). The original packaging, installation position, customer markings, usage duration, and failure phenomena must correspond one-to-one with the failed units.
• Criteria: Failed units with missing original evidence (photos, packaging, serial number records) can only be used as reference samples and cannot serve as the basis for conclusions.

Step 2: Non-Destructive Analysis, Identify Suspicious Areas.

Visual (10× to 50× magnification) → X-ray (solder voids, cold solder joints, internal structure) → CT if necessary. The output is a "list of suspicious areas."

• Criteria: Solder void area ratio >25%, crack length exceeding 30% of the solder joint width, and foreign object diameter >0.1 mm must be listed as suspicious items.
• If the conclusions from the three non-destructive methods are consistent, proceed directly to electrical performance verification; if inconsistent, retain the samples and do not prematurely decapsulate.
• Do not alter the state of the failed units; prohibit any decapsulation actions before this step is completed.

Step 3: Electrical Performance Verification and Failure Reproduction.

Test each item according to the specification book and categorize the results into "complete failure," "parameter drift," and "intermittent failure."

• Criteria (NFF): If the proportion of intermittent failures, or "no fault found" units, in a single batch is >20%, the case cannot be closed as "passed testing." Additional environmental stress (temperature cycling −40 ℃ to 85 ℃, 5 cycles) or continuous vibration for 2 hours must be applied to reproduce the failure; if still unable to reproduce, record it in the "unreproducible list" and classify by phenomenon.
• Reproduction Rate Criteria: If the failure can be consistently reproduced (at least 3 times in a row) → proceed to mechanism localization; if intermittently reproduced → limit the conclusion to "condition-related"; if completely unreproducible → the deliverable can only be a probability distribution and stress correlation conclusion.

Step 4: Destructive Localization, Lock in the Mechanism Level.

Decapsulation and delayering → cross-sectioning → microscopic observation and elemental analysis. The criteria are "the mechanism must be identified at a named level":

• Criteria: Do not output the root cause before the mechanism level is determined; compare the cross-section samples with non-failed units from the same batch to confirm unique characteristics of the failed units.
• Each level must provide an "exclusion list," detailing which mechanisms were excluded and the basis for exclusion.

Step 5: Root Cause Closure and FMEA Linkage.

Map the mechanism to adjustable process parameters or usage conditions, and verify by reproducing and eliminating the failure under the same stress conditions.

• Verification Criteria: After implementing corrective actions, the reproduction rate must be 0/10 units under the same stress conditions, and there must be no similar failures in 3 consecutive batches (or a cumulative total of 500 units) to close the case.
• Linkage Requirements: Confirmed failure modes must update the failure effects, failure causes, and detection measures in the PFMEA, and the corresponding parameters must be added to the control plan; if design-related, transfer to DFMEA and the design change process.
• Time Limit: Issue a preliminary report (phenomenon + localized mechanism + interim measures) within 48 hours, and complete the full FA report within 10 working days.

The FA report must have a fixed structure (missing any section is considered incomplete): ① identification and usage information of the failed units; ② failure phenomena and reproduction conditions; ③ results of non-destructive and electrical testing; ④ images and conclusions from destructive analysis; ⑤ mechanism determination and root cause (including exclusions); ⑥ corrective actions and verification results; ⑦ linkage records with FMEA/control plan; ⑧ handling and statistics of unreproducible units.

4. Common Pitfalls

Pitfall 1: Dismantle Before Investigate. The most common and costly mistake. Preemptive destructive actions destroy on-site information in one go, leaving all subsequent conclusions to speculation. The rule is: do not act unless absolutely necessary, and always ask, "Is there a less destructive observation method?"

Pitfall 2: Single Unit Root Cause Determination. Concluding from a single failed unit without batch comparison or non-failed unit reference cannot rule out randomness. Sample quantity and retention are the minimum requirements for conclusion strength.

Pitfall 3: Treating "Component Defect" as the Root Cause. Component defects are just a level of the phenomenon; the question "why does this component fail under these conditions" is the true root cause. The component chain and process chain are two independent evidence chains and cannot substitute for each other or be concluded under the supplier's name.

Pitfall 4: Judging "No Problem" for NFF. "No fault found" is a signal of insufficient test resolution and stress conditions, not evidence that the failure does not exist. Returning NFF units to the customer one by one is equivalent to voluntarily giving up the most valuable stress information.

Pitfall 5: FA Report Only States Conclusions, Not Exclusions. If the report does not record "which mechanisms were excluded and why," future analysts cannot reuse the reasoning path, and the same issue will have to be addressed from scratch the second time it appears.

5. Self-Check List

1. Were the returned units numbered, photographed, and sealed before disassembly, with ≥2 undamaged samples retained?
2. Was non-destructive analysis conducted before electrical and destructive methods? Were samples retained for verification when the three non-destructive conclusions were inconsistent?
3. Is the mechanism level clearly identified as material, process connection, electrical stress, or mechanical/environmental, with batch comparison evidence?
4. Is the root cause linked to adjustable process parameters or design parameters, and have they been updated in the PFMEA and control plan?
5. Is the closure criterion based on a reproduction rate of 0/10 units and no similar failures in 3 consecutive batches?
6. Were the preliminary report issued within 48 hours and the complete report completed within 10 working days?

Preserve evidence before disassembly; early disassembly means guessing the cause.

Knowledge code: 10.2.2

Version: v20261014

Author: QTank QTank is dedicated to providing systematic professional knowledge, methodologies, and practical tools for quality management practitioners, helping companies continuously improve their quality capabilities.