BGA and QFN X-Ray inspection reports are useful only when a buyer can tell what product was examined, which hidden features were in scope, how the images were acquired, what was measured, which acceptance source controlled the decision, and what the method could not establish. A page containing one grayscale image and a green PASS label is not a complete release record.
X-Ray uses differences in material attenuation to create an image of internal geometry and density. It can reveal selected bridges, opens, voiding patterns, insufficient or excessive solder, misalignment and other structural clues in hidden-joint packages. However, a two-dimensional projection compresses three-dimensional structures, overlapping features can obscure each other, and image appearance changes with voltage, current, magnification, angle, detector, filtering and analysis settings.
This buyer guide does not repeat a production sampling or acceptance-plan article. Its single task is to explain how to read the delivered report, test whether the evidence supports the stated disposition, and decide when complementary inspection or escalation is required.
The review should also distinguish a documented observation from an engineering inference, because the same image pattern can have multiple causes.
Confirm each BGA and QFN X-Ray report matches the product
Start with identity. The report should name the customer or project, board part number and revision, assembly variant, work order or lot, panel and unit or serial where required, side, package reference designator, device type, date/time, equipment, program or recipe revision and operator or reviewer. If the report uses sample units, it should state how they were selected and which population the result represents.
Confirm orientation. A BGA array, QFN thermal pad and perimeter can be misread when the image lacks a reference corner, board view or annotated designator. The report should connect each image to a component and unit, identify top/bottom orientation, and mark the region under review. File names such as IMG0007 are not enough unless the report maps them unambiguously.
Check the governing configuration: PCB and component package, land pattern, stencil and aperture where relevant, solder paste/alloy, side and reflow pass, component source, rework status and criteria revision. An image from an earlier engineering build does not release a later configuration unless the project has documented why it remains representative.
An X-ray image is auditable only when it is linked to the board, revision, recipe, view and disposition. Read the acquisition method before reading the defect
Image quality determines what can be interpreted. Review equipment type, 2D/2.5D/3D or CT capability, tube voltage and current where retained, magnification, pixel or voxel size where relevant, detector, angle, image averaging, filters, contrast and analysis recipe. Not every production report needs every physics parameter on its face, but the controlled recipe and equipment identity must be retrievable.
Magnification is not the same as usable resolution. A large image can still be blurred, saturated or noisy. Look for clear joint boundaries, stable contrast, minimal motion, appropriate field of view and visible reference features. If dense heat sinks, shields, connectors, copper planes or double-sided components overlap the region, an angled view or CT reconstruction may be required.
Compare images only when the acquisition is comparable. Automatic thresholds used to segment solder and void regions can change measured percentages. A supplier should control the algorithm and settings, identify manual overrides, and validate the recipe against known structures. An attractive color map is not a substitute for a repeatable measurement method.
NASA’s public hidden-solder inspection material illustrates both the value and limitations of X-Ray for BGA evaluation. It should be used as technical context, not as an automatic acceptance contract for a commercial product. Project drawings, purchased standards and approved criteria remain authoritative.
Interpret BGA evidence by location and pattern
A BGA review should identify the array, corners, edges, center and any thermally or electrically critical joints. Look for joint presence, relative size and shape, alignment, bridging, apparent opens, solder distribution, head-in-pillow indicators where the method can support them, and void location or pattern. Compare with a known-good reference and across the same package population rather than judging one ball in isolation.
Patterns can be more informative than a single anomaly. A row shift may suggest placement or registration; asymmetry can point to coplanarity, warpage or thermal imbalance; repeated void patterns can connect to paste, via design or reflow; a localized bridge may relate to excess deposit or collapse. These are hypotheses, not final causes. The report should separate the observed image feature from the proposed mechanism.
A standard 2D projection may not reliably prove an interface crack or distinguish every open mechanism. Apparent circularity does not demonstrate metallurgical wetting. If the electrical symptom, process data and image disagree, escalate to angled views, CT, electrical localization, cross-section or another method selected for the question.
Multiple views can separate real BGA joint conditions from overlap, projection and imaging artifacts. Question void measurements before accepting the number
A void percentage requires a defined numerator, denominator and segmentation method. Ask whether the measurement applies to one ball, an average, a maximum, a thermal region or the full sample. Confirm whether overlapping copper or via structures affect the projected area, how thresholds are selected, whether edge voids and clustered voids are distinguished, and which product criterion controls the result.
Do not convert a generic void number into a universal quality score. Location, geometry, joint function, thermal and mechanical loading, reliability requirement and applicable specification matter. Report raw or per-location values where the buyer needs trend analysis, and retain images for outliers. A lot average can hide one critical joint.
Interpret QFN perimeter and thermal-pad evidence separately
QFN packages combine perimeter terminations with an exposed thermal or ground pad. The X-Ray report should state which features it can evaluate and how overlapping package metallization, PCB copper, vias and solder affect the image. A thermal-pad void calculation does not automatically evaluate every perimeter connection, and visible perimeter geometry does not prove the thermal interface is acceptable.
For the exposed pad, review solder distribution, large continuous voids, peripheral escape paths, via effects, excessive or insufficient deposit and package alignment. For perimeter joints, review location, symmetry, solder presence and possible bridges or opens within method capability. Side fillet expectations depend on package termination design, land pattern and acceptance source; they should not be invented from the X-Ray image.
Compare the reported pattern with stencil segmentation, aperture, via design and the released reflow route. QFN voiding can be influenced by paste deposition, outgassing paths, pad design, via construction, component finish and thermal profile. The report can expose the pattern but should not declare a process root cause without corroborating print, reflow, material and cross-sectional evidence.
QFN void statements need a defined region of interest, measurement method and acceptance basis. Check criteria measurements and reviewer decisions
The report should cite the controlled source of acceptance: customer drawing, product specification, approved workmanship standard, package or component requirement, validated internal rule or deviation. It should identify the revision and precedence when sources conflict. “IPC compliant” is incomplete when the document, revision, class and applicable feature are not stated.
For each measurable characteristic, require the unit, region, algorithm, limit and result. For qualitative review, define defect categories, image examples or reviewer guidance and escalation. Record the original automated result, manual review, any override, reason and authority. If the image cannot support a decision, the correct output is insufficient evidence or escalation—not an assumed PASS.
Sampling should name the population, selection method, package scope, view count, frequency, acceptance rule, response to one failure and return-to-normal condition. An NPI review may require wider coverage than stable production; a process change or recurring defect can reopen the plan. The related GNS article on BGA X-Ray acceptance planning explains how to build that release route.
Connect the report to process and complementary evidence
X-Ray should sit inside a wider evidence chain. Link findings to solder-paste print and SPI data, component and PCB identity, placement program, reflow profile, AOI or microscope results, electrical/functional test, repair, retest and final disposition. A repeated image signature becomes more useful when it can be correlated with a material lot, process window, program change or failure mode.
Use complementary methods according to the unresolved question. AOI or microscopy can review accessible side features and neighboring placement. ICT, flying probe, boundary scan or FCT can localize an electrical symptom. CT or cross-section can investigate geometry or interfaces. Thermal, mechanical or environmental validation can address product-level risk. No method should be added merely to create more pages in the report.
The GNS explanation of AOI, X-Ray, ICT and FCT helps buyers assign each method to the evidence it can actually provide. The supplier should state coverage and exclusions rather than implying that all hidden-joint failure mechanisms are detected.
X-ray findings should be interpreted with package context, visible evidence, test results and disposition history. Use a buyer report acceptance checklist
Before accepting a report, confirm five groups of evidence: identity, acquisition, coverage, decision and follow-through. Identity ties every image to the exact assembly and feature. Acquisition shows a controlled and adequate recipe. Coverage states packages, locations, views and sampling. Decision connects observations and measurements to criteria and reviewer authority. Follow-through preserves defects, holds, rework, retest, process response and final release.
Reject or return reports with unlabeled images, missing orientation, unknown recipe, no criteria source, unexplained PASS labels, only lot averages, missing sample definition, manual overrides without reason, or conclusions that exceed the method. Ask for raw or full-resolution evidence where compression hides features, and require controlled retention rather than relying on a temporary screen.
When a report reveals a new pattern, the response should identify the affected population and production window. Confirm containment, review related process inputs, choose targeted follow-up and preserve the original data. A corrected image does not close a process excursion until product disposition and recurrence controls are documented.
Compare NPI and production reports consistently
During NPI, require a baseline set that covers representative packages, corners, thermal extremes, process variants and known areas of concern. Retain the full-resolution images, acquisition recipe and annotated measurements as a controlled reference. Production images should be compared only when product construction, equipment and recipe remain sufficiently equivalent. If the supplier changes magnification, angle, detector, filtering or segmentation, validate the new method before treating trends as continuous.
Create defect codes that describe observations rather than assumed causes: possible bridge, low apparent solder, abnormal void cluster, misalignment, incomplete image coverage or measurement invalid. Link each code to review guidance and escalation. A root cause such as stencil aperture, placement offset or profile imbalance should be added only after corroborating evidence. This prevents automated dashboards from turning early interpretations into permanent facts.
Ask the supplier to demonstrate a report for a failing or challenged sample as well as a normal board. The demonstration should show how the unit enters hold, how manual review is recorded, whether raw images are preserved, how rework and reinspection remain linked, and how a lot-level response is triggered. A system tested only with passing images may have an unverified exception path.
Review measurement repeatability when numeric limits drive release. Have the same feature measured repeatedly, by different reviewers where manual input matters, and after routine loading or recipe restart. Investigate threshold sensitivity and edge cases. If measurement variation consumes a large share of the product limit, improve the method or create a guarded review band rather than reporting false precision.
For customer reports, decide whether every image, only exceptions, representative samples or a summary is needed. Preserve detailed records at the supplier even when the customer receives a compact pack. State retention period, retrieval time, file format, naming and access. Screenshots pasted into a PDF may be acceptable for review only if the underlying traceable source images remain available.
Finally, connect X-Ray trends to process engineering without allowing automatic process adjustment to bypass approval. A repeated shift or void pattern can trigger print, placement, material and profile review. Any resulting stencil, recipe or criteria change should identify the affected configuration, validation evidence and effective production window.
Include an image-quality failure code. When the region is obscured, saturated, blurred or outside validated resolution, classify the result as invalid and repeat or escalate the inspection instead of recording “no defect found.”
Conclusion
A useful BGA or QFN X-Ray report is a traceable engineering record, not a gallery. It identifies the product, package and orientation; records the acquisition method; defines scope and criteria; distinguishes observation from inference; reports measurements and limitations; and connects every exception to disposition and process response.
Before quotation or NPI, send the package list, critical locations, drawings, acceptance sources, population, sampling intent, required measurements, report fields and escalation rules. Ask the supplier to provide a sample report and demonstrate retrieval. This lets the OEM evaluate evidence quality before hidden-joint results become a shipment dispute.
Request an X-Ray Evidence Review
FAQ
Can one X-Ray image prove that every BGA joint is acceptable?
No. One projection may show selected density and geometric features, but overlapping structures, view angle, resolution, recipe and interpretation limit what is visible. Define coverage, views, criteria and complementary evidence.
Does BGA void percentage alone determine acceptance?
No. The calculation method, segmentation threshold, projection overlap, joint or region identity, product criteria and functional or reliability significance must be defined. A percentage without method and location is not a release rule.
What should a QFN X-Ray report include?
Include board and package identity, orientation, image and recipe details, views, thermal-pad and perimeter-joint scope, measurement method, acceptance source, images tied to locations, results, reviewer, disposition and limits of detection.
When should X-Ray findings be escalated?
Escalate new or repeated patterns, possible opens or bridges, unexplained asymmetry, shifted packages, out-of-limit measurements, image-quality failures, criteria conflicts and findings that may indicate a wider print, placement, reflow or material problem.
