Investigating MLCC Cracks: Evidence Before Root Cause

Investigating MLCC Cracks: Evidence Before Root Cause
On this page

A multilayer ceramic capacitor (MLCC) can develop a crack during its life in an assembly, but finding a cracked capacitor does not by itself show which operation caused it. Depaneling, a test fixture, enclosure assembly and thermal exposure are among the possibilities to investigate. The useful first step is to separate what has been observed from what the team suspects.

This guide follows a fictional investigation from a low-insulation-resistance result to a plan for comparing evidence. For the broader choice of separation method and fixture support, see the PCB depaneling guide. For fixture access and loading during in-circuit test, see ICT test-point planning.

Start with the actual observation

In the example, final electrical test finds low insulation resistance on a power rail of an assembled board. Two MLCCs on that rail, C17 and C18, sit near a panel break. The board has also passed through a bed-of-nails test fixture and been installed in an enclosure before the final test. The team does not have a comparable insulation-resistance result from immediately before each operation.

The finding is low insulation resistance on the rail, not yet a confirmed crack in either capacitor. A rail measurement alone does not identify which component or other path accounts for it. The quality owner holds the affected material according to the applicable procedure and asks for the test conditions and results to be retained with the board identity and revision.

Agree an investigation and specimen plan with the responsible team and a qualified failure-analysis provider, including whether and how electrical evidence can help locate the leakage path. The plan should account for the condition and preservation of the available specimens.

The location of C17 and C18 makes depaneling worth examining. It does not eliminate pressure from the test fixture, mounting loads, thermal exposure or a component-related issue. Record these as competing hypotheses until evidence narrows them.

What should the team preserve and compare?

Keep each suspect assembly identifiable and record what is known about its handling after the failure was found. Avoid an unrecorded rework or component replacement that could remove evidence needed by the investigation. The responsible quality owner and failure-analysis provider should agree on the disposition and handling of specimens under the applicable procedures.

Collect a comparison group whose history is understood as well as possible. A board that passed the same test may be useful for comparison, but a pass does not prove it is free of a latent crack. Record differences in revision, part number and lot, production route, fixture, rework history and test conditions; otherwise an apparent difference may have another explanation.

Evidence What to look for What a gap means
Electrical result Board identity, rail, test method and conditions, result and any earlier comparable measurements Without comparable checkpoints, the first operation after which the result changed is unknown.
Component and assembly Exact MLCC part numbers, lots, reference locations, orientation, board revision and visible condition Similar locations or markings do not establish identical parts or exposure.
Process route Sequence, equipment and fixture identities, support arrangements, relevant settings, changes and exceptions An assumed route may not match the path the board actually took.
Post-test handling Rework, removal, transport and any additional tests after discovery Later handling can complicate interpretation of the original condition.
Comparable assemblies Test results and known differences for boards from the relevant build A passing comparison is useful context, not proof that all other parts are intact.

The depth of records available varies by build. Mark an unavailable result as unknown; do not reconstruct it from a later pass or from the intended route. The broader traceability guide explains how agreed records help identify relevant builds and material lots.

Which operation should be investigated first?

Start with the operations that could have loaded the suspect region, while keeping the rest of the route open. In the example, the sequence below is an investigation map, not a ranking of the likelihood of each cause.

Hypothesis Records or observations to examine A finding that would guide the next step
Board loading during depaneling Panel drawing, separation method, fixture and support, component position, equipment changes and actual sequence A change in support or separation setup affecting the suspect build warrants a focused comparison under representative conditions. It is not proof of causation.
Test-fixture loading Fixture revision, probe and support locations, setup changes and test records If the fixture or support differs between affected and comparison boards, involve the fixture owner in evaluating the loading at the suspect region.
Mounting or connector force Enclosure and support geometry, assembly sequence and recorded exceptions A mismatch or interference is a reason to review the actual mounting operation and its loading, not to assign responsibility from proximity alone.
Thermal or component-related damage Part identity and lot, assembly and rework history, relevant temperature records, supplier information and physical analysis The component supplier and analysis provider can help assess whether the observed evidence is consistent with a particular mechanism and what alternative explanations remain.

The Murata explanation of strain cracks describes how board bending can transfer stress through solder joints to a ceramic capacitor. Its example is useful for understanding a possible mechanism; it is not a record of the fictional assembly or proof that a panel break caused its failure.

What can electrical tests and a cross-section tell you?

Electrical evidence can show whether a measured property has changed under the test conditions. TDK identifies lower insulation resistance, increased leakage, loss of capacitance and increased dissipation factor as possible indicators of a flex crack. These indicators need interpretation alongside other evidence: they do not identify the operation that caused damage, and a passing result cannot rule out a latent crack.

Visual inspection also has limits. A capacitor may look normal from outside even when a crack is present. TDK describes cross-sectioning as a common way to find a flex crack, and Murata publishes an example of a polished section with a diagonal crack near a termination. Both sources also frame the damage in terms of mechanical stress, but a photograph or an angle alone cannot determine the event that produced a particular crack.

If analysis is needed, agree the question, specimen history, comparison material and analysis plan with a qualified failure-analysis provider before destructive work. Ask the provider to report what was observed, the limits of the examination and alternative explanations. A single section represents what was examined in that specimen; an unobserved crack elsewhere cannot be excluded solely by that result.

How should the team test a suspected cause?

Suppose the lab reports a crack consistent with mechanical loading and the records show that the depaneling fixture changed for the affected build. That combination makes the fixture change a testable hypothesis. It does not establish that the change caused the crack, especially if fixture testing or enclosure assembly also loads the area.

The process owner can define a controlled comparison using representative board construction, population, tooling, supports and operation sequence. Where strain measurement is appropriate, agree the measurement locations, method, interpretation and project-specific acceptance basis with the responsible engineering and quality teams before testing. Compare the actual setup with the proposed change and retain the results with the configuration tested.

If a strain measurement shows a reduction after the change, it provides evidence at the measured locations under those conditions. It does not by itself prove that every capacitor is undamaged or that the final product is qualified. If the evidence does not distinguish the competing hypotheses, record the result and decide what additional examination or process comparison is justified.

What should be in the investigation handoff?

A useful handoff gives the component supplier, assembler or analysis lab enough context to work from the same facts:

  • The observed symptom, measurement conditions, affected board identifiers and revision.
  • Suspect reference designators, exact part numbers and lots, plus their locations on the assembly.
  • The actual production and handling sequence, known changes, rework and missing records.
  • Available failed and comparison specimens, their test and handling histories, and any restrictions on destructive analysis.
  • The competing explanations, the decision the analysis should inform, and the responsible contact for disposition.

Keep the conclusion proportional to the evidence. A report can say that a crack was observed, that a loading mechanism is consistent with the evidence, or that a process change reduced measured strain under stated conditions. Naming a particular operation as the cause requires the additional evidence needed to distinguish it from the alternatives.

APTPCB offers PCB fabrication and turnkey PCB assembly. For a PCB or assembly quote, share the board and assembly files, component requirements and any handling, inspection or test expectations. Confirm the investigation or analysis scope and deliverables for the project before relying on a particular service.

Frequently asked questions

Does a diagonal crack prove the board was damaged during depaneling?

No. TDK describes a diagonal pattern associated with flex cracking, but several operations can bend a board and other crack patterns are possible. Compare the physical findings with the actual process history and other evidence.

Can a capacitor pass electrical test and still have a crack?

Yes. A crack may not produce a detectable change under a particular test's conditions. A pass is a result for that test, not a guarantee that the ceramic body is undamaged.

Should every suspect capacitor be cross-sectioned?

Cross-sectioning is destructive, so agree the investigation question and specimen plan with the analysis provider and quality owner. A single section cannot establish the condition of every component in a build.

References