BGA Inspection: Match X-ray and Electrical Evidence

Separate visible placement from hidden connections

Package alignment, orientation, surrounding components and visible contamination can often be reviewed optically. Most BGA solder interfaces cannot be inspected directly from above. Oblique views or specialized optical access may help at package edges but do not establish the condition of all interior joints.

The AOI page owns the general optical process. Here the important question is which BGA features remain hidden and which complementary method can resolve them.

Choose X-ray views for the suspected condition

Two-dimensional projections superimpose structures along the beam path. They can show useful geometry, but overlapping opposite-side components and subtle interface separations may complicate interpretation. Angled views or suitable three-dimensional techniques may be needed for a specific question.

Define resolution, orientation, region of interest and the actual feature to evaluate with the inspection specialist. A machine model or a magnification number alone is not a detection guarantee. Retain the image settings with the specimen identity.

Define void and shape measurements carefully

If the requirement concerns projected void area, specify the joint boundary, segmentation method and whether the reported value is per joint or an aggregate. A two-dimensional area fraction is not automatically a three-dimensional volume fraction. Different thresholds can produce different reported areas from the same image.

Acceptance limits must come from the applicable controlled criteria and component or customer requirements. This page supplies no universal acceptable void percentage. The tool counts observed affected joints in a declared sample; it does not analyze images or decide conformity.

Correlate with electrical and process records

An open circuit may be intermittent or insensitive to one test state. A joint can conduct during a static check yet remain unsuitable under thermal or mechanical stress. Conversely, a suspect image needs interpretation rather than automatic rejection based on a visual resemblance.

Associate X-ray findings with the same unit's electrical log, paste inspection, placement history and reflow record. The X-ray inspection guide explains the broader method boundary without duplicating this package-specific review.

Observed affected-joint fraction

Preserve failures before destructive analysis

Record package, ball coordinate, board serial, orientation and symptoms before rework. Removing or reheating a component can alter the evidence. If a specialist proposes cross-sectioning or another destructive method, agree the sample and location first.

Keep unaffected controls when available. A cut through one joint cannot establish the state of all joints, and an image alone cannot assign the entire root cause. Distinguish the observed separation from the process mechanism believed to have produced it.

Use rates with a declared denominator

For an illustrative sample of 400 evaluated joints and eight affected joints, the observed affected-joint fraction is two percent. That is different from the fraction of boards containing a defect. The result also depends on the selected inspection method and sample selection.

Report repaired and retested units separately from first-pass results. A low observed fraction in a small or selectively chosen sample is not a claim of production reliability. Preserve the unresolved question if detection capability has not been demonstrated.

Request a BGA inspection scope that answers the question

Provide the package drawing, assembly revision, affected region, previous processing, observed symptom and required acceptance basis. State whether the work concerns first-article review, production screening or failure analysis.

If the goal is finding a repeatable mechanism, continue to the PCB failure-analysis route. RAYPCB can review the requested manufacturing and inspection scope from the actual project files; equipment availability and reporting requirements should be confirmed in the proposal.

Questions to resolve before the next build

Is an X-ray image with uniform balls enough for release? Uniform appearance is one observation under a particular imaging setup. The acceptance question may also involve hidden interfaces, electrical behavior or features that the chosen projection cannot resolve. Request the applicable criterion, image settings and complementary evidence. Avoid presenting a good-looking example image as proof that an entire production lot received the required inspection.

Should every suspicious joint be reworked immediately? Preserve its location and relevant electrical evidence first, then follow the approved disposition. Reheating can change the feature that an investigation needs to examine. If the issue is recurrent or unexplained, retain suitable samples for analysis and separate containment from root-cause work. The correct sequence depends on the project risk and available evidence, not only on whether repair appears quick.

Make inspection evidence useful to the next decision

Start with a feature map that identifies what the station is expected to observe. Use actual component references, pad or joint locations and product variants. A statement that the board receives inspection is too broad to reveal whether a critical hidden or shadowed feature is included. Mark exclusions and the complementary method intended to address them.

Define how images or numerical results are associated with the unit. Board identity, panel position, program revision and process stage are useful only if they remain connected. An image exported without those references may be difficult to interpret later, particularly when the product has several similar variants or the program changed during the run.

Review the measurement definition before comparing data. Area, height, volume, position and an operator's defect classification are different quantities. A count of calls is not a count of defective boards. If two stations or two reports use different boundaries, normalize the definition or keep the results separate rather than averaging incompatible numbers.

Establish the review workflow for calls and uncertain observations. Record confirmed findings, false calls, invalid measurements and unresolved cases distinctly. Give the reviewer access to the applicable criteria and enough context to make the decision. A busy production line should not force the operator to choose between pass and fail when the actual result is not yet interpretable.

Validate changes using representative conditions. A new library, threshold, lighting arrangement or fixture can alter detection. Compare against a reviewed set of observations and retain the change record. Fewer alarms may improve efficiency, but only if the required detection behavior is preserved. Likewise, more alarms do not automatically mean better quality.

Connect findings to the earliest useful process evidence. A defect found after reflow may need paste and placement information from the same unit. A hidden-joint concern may require imaging and electrical results together. Keep the investigation specific enough to test a hypothesis rather than assigning every defect to a generic machine setting.

When requesting an inspection service, state the purpose: setup review, routine screening, sampling or investigation. Specify the report detail and retention needed for the decision. A thumbnail image and a result flag may be adequate for one task but insufficient for another. Agree that difference before the quotation is finalized, and keep any unresolved capability or acceptance question separate from the production release.