Sunking PCB

PCB Failure Analysis: Choosing Methods in the Right Order to Reach a Defensible Root Cause

Design Guide2026-10-23Sunking Circuits Technical Team
PCB Failure Analysis: Choosing Methods in the Right Order to Reach a Defensible Root Cause

Failure analysis is a sequence of questions, not a single test

When a board fails in test or in the field, the reflex is often "send it for cross-sectioning." But a microsection answers exactly one question: what does the structure look like on this particular cut plane. If the cut is in the wrong place, or the failure site was never confirmed before cutting, the report can be perfectly sharp and still support no conclusion at all.

A more useful approach treats analysis as a chain of questions, each answered by one step:

  1. What type of failure is it, and can it be reproduced reliably?
  2. Where is it? Can it be narrowed to one net, one hole, one solder joint?
  3. Is the structure at that location abnormal, and how?
  4. At which stage did the abnormality form: fabrication, assembly or use?
  5. Do boards from the same lot that did not fail show the same feature?

Each question has a matching method, and the answer to one question determines how the next method should be applied. Run them out of order and the later evidence loses its meaning.

Classify the symptom before choosing where to start

Before requesting any analysis, write the symptom as one sentence and place it in one of these groups. The classification itself decides the first step.

  • Open, short or abnormal resistance: confirm electrically first and narrow the problem to a specific net and location.
  • Intermittent failure (appears with temperature, vibration or flexing): try to reproduce it under controlled conditions and record the trigger. Destroying the sample early often removes the only clue.
  • Delamination, blistering, measling: record location and distribution first, then judge whether it correlates with a thermal process. See lamination void and delamination prevention.
  • Solderability or joint defects (non-wetting, weak joints, cracked joints): separate the pad side from the component side first, then look at surface finish, storage and reflow. Common joint-side causes are covered in PCBA solder joint failure causes.
  • Dimensional, warpage or cosmetic deviation: check against the drawing and the agreed acceptance standard before discussing causes.

The value of this step is that it forces whoever requests the analysis to state what they actually want to know. Many reports end up unusable not because the method was wrong, but because the question was never defined.

What each method can prove, and what it cannot

MethodCan answerCannot answerDestructive
Electrical test (continuity, isolation, resistance)Which net failed, failure typePhysical cause, stage of originNo
Visual and stereo-microscope inspectionVisible surface defects, location, distributionInternal structure, cross-section conditionNo
X-ray inspectionInternal voids, misregistration, foreign materialMicroscopic interface condition, chemistryNo
MicrosectionPlating, interfaces, cracks, separation on one planeAnything outside that plane, time of formationYes
Surface and composition analysisPresence of contamination, oxidation, unexpected elementsWhen and where they were introducedDepends on method
Reproduction test (thermal stress, reflow simulation)Whether the failure recurs under a defined stressWhether field stress matches the testUsually

The most important column is "cannot answer." Finding the open net does not explain the cause. Seeing a crack in a section does not tell you whether it was there at shipment or formed in service. Treating "observed" as "root cause found" is the most common logical jump in failure analysis.

Method definitions and procedures follow the IPC-TM-650 test methods manual, while visual and cross-section acceptance criteria build on IPC-A-600. For an overview of routine inspection methods, see PCB quality inspection methods.

Non-destructive first, destructive last

Sectioning, peeling and de-layering cannot be undone. That gives the sequence three hard constraints:

  • Locate before you cut. Once electrical test and X-ray have narrowed the failure to one hole, one trace segment or one joint, the section has a defined target. Cutting without localization is random sampling across a large area; hitting the defect becomes luck.
  • Record before you touch. Photos, electrical readings and failure conditions must be captured before any cleaning, rework or component removal. A reworked board has a rewritten thermal history at the joint and pad, and what you see afterwards is hard to attribute.
  • Reproduce before you consume. If an intermittent fault can be reproduced under control, the trigger is confirmed. Cut the sample first and you lose the reference you need to verify that a corrective action works.

A practical check: if the next step will destroy the sample, ask whether any non-destructive method can narrow the scope a little further. Usually one can.

Preserving samples and context

How far a conclusion can be trusted depends heavily on what was sent with the sample.

Sample set:

  • The failed board itself, in its failed state: not cleaned, not reworked, no components removed.
  • Boards from the same lot that did not fail, as a reference. Without a reference, you cannot tell whether a feature is abnormal or simply how that lot was built.
  • Unassembled bare boards from the same lot, if any remain, to separate fabrication from assembly.
  • Lot and date-code markings where traceable; marking rules are covered in PCB traceability marking design.

Context:

  • Stage at which the failure appeared: incoming, post-assembly test, burn-in, field use.
  • Failure proportion and distribution: concentrated in one lot or one panel position, or random.
  • Assembly and use conditions: number of reflow passes, rework history, storage and moisture handling (see PCB moisture control, packaging and storage), operating environment.
  • Checks already performed and their results, so samples are not consumed twice.

Without reference boards and context, an analysis lab can usually only describe what it observed. It cannot answer who owns the problem.

Three verdicts: fabrication defect, design or material limit, assembly or use stress

A conclusion you can act on assigns the cause to one of three groups and states the evidence:

  • Fabrication defect: the abnormal structure is also present on bare or reference boards and departs from the agreed acceptance standard, for example plating, hole wall or interlayer bonding that is already abnormal before assembly. This leads back into the fabrication process; the relevant inspection steps are covered in PCB drilling hole quality inspection and electroplating quality inspection and release.
  • Design or material limit: the board was built to drawing and standard, but a design feature or material choice cannot withstand the real operating stress. Normal reference boards plus failure only under a specific stress usually point here. The fix is a design or material change, not a fabrication claim.
  • Assembly or use stress: bare and reference boards are normal; the abnormality appears only on boards that went through a specific assembly process or use condition. Rework, thermal exposure beyond what was agreed, mechanical stress and humid environments belong here.

The groups often overlap, for instance a material limit amplified by an assembly thermal process. The conclusion should then name the primary cause and the contributing factors, not settle for "multiple factors." For the acceptance boundaries themselves, see PCB defect classification and Class 2 vs Class 3 acceptance.

What to state when requesting an analysis

Whether the request goes to the board fabricator, an independent lab or your own quality team, it should state:

  • The question to answer, in one sentence, such as "confirm the physical location and stage of origin of the open on this net."
  • Symptom class and reproduction conditions, using the classification above and any known trigger.
  • Sample list: how many failed, reference and bare boards, with markings and condition.
  • Permitted destruction: which samples may be sectioned and which must be retained.
  • Acceptance basis: the agreed standard and class; see tolerances and acceptance terms.
  • Expected output: an observation report only, or a cause classification with improvement recommendations.

The fabricator-side inspection system and process controls are described under quality, and the buildable range under capabilities. To discuss a specific failed sample, background information can be submitted through quote and engineering inquiry.

FAQ

Q: Can a failed board go straight to microsectioning?

A: It can, but confirm the failure site first. Without electrical or X-ray localization, the section is taken at random across a large area. If it misses the defect, the report supports nothing and the sample is already gone.

Q: A crack is visible in the section. Does that prove a fabrication fault?

A: Not on its own. A crack can form during fabrication, reflow, rework or service. Compare with bare or non-failed boards from the same lot; only if the reference samples show the same abnormal structure does the evidence lean towards fabrication.

Q: Why send non-failed reference boards?

A: Without a reference, there is no way to tell whether a feature is abnormal or normal for that lot. The reference is what turns a description into a cause judgment.

Q: Can a reworked board still be analysed?

A: Yes, but the conclusion is limited. Rework rewrites the thermal history of the joint and pad, so solder-related structures are hard to attribute to original fabrication or assembly. Where an unreworked failed sample exists, send that one first.

Q: How do you analyse an intermittent failure?

A: Try to reproduce it under controlled conditions first, record the trigger such as temperature, vibration or flexing, and localize it electrically while it is in the failed state. Avoid destructive steps until reproduction and localization are done.

Q: Is "caused by multiple factors" an acceptable conclusion?

A: It is better not to accept it. The conclusion should place the primary cause in fabrication defect, design or material limit, or assembly or use stress, and name contributing factors with evidence. Only a clear primary cause gives the corrective action a clear owner.

Need Professional PCB Manufacturing?

Founded in 2009, we provide one-stop PCB solutions from prototype to mass production.