Computed Tomography for PCB and Solder Joint Analysis

Conventional X-ray inspection produces a shadow image, and everything in the beam path is superimposed. A BGA solder ball on the top side hides the joint on the bottom side, and a plated barrel is flattened into a dark line. Computed tomography solves that by reconstructing the object in three dimensions. This guide explains how gopcb uses CT for PCB and solder joint analysis, what it can resolve, and where its limits lie.

What Computed Tomography Adds

Computed tomography takes a series of X-ray projections from many angles and reconstructs them into a three dimensional volume. The result can be sliced at any plane, so an inspector can follow a single solder joint through its full height rather than inferring its shape from a flat image. Hidden structures become directly visible. A joint that looks acceptable in projection can be shown to have a crack running through half of its interface.

That capability matters most where joints are concealed, where several layers of metal overlap, and where a failure has to be understood rather than simply detected. Detection can be done with a two dimensional system, but understanding a defect mechanism usually requires the third dimension.

How the Scan Works

The sample rotates while the source and detector stay fixed, and hundreds or thousands of projections are recorded. A reconstruction algorithm converts that projection set into a volume dataset, which is then displayed as slices, as a surface rendering or as a cutaway. Scan time grows with the number of projections and the required resolution.

Geometric magnification is what delivers resolution. Placing the sample close to the source enlarges the projection on the detector, but it also limits how much of a large board can be scanned in one pass. Large panels are therefore scanned region by region, with each region reconstructed separately. Stitching those regions into one dataset is possible, but it adds time and introduces alignment error at the boundaries.

Resolution and Contrast Limits

Resolution depends on the source spot size, the detector pixel size and the magnification achieved. Microfocus sources reach a few micrometres, which is enough to resolve a solder ball and its internal voids, but not enough to image intermetallic layers, which require electron microscopy.

Contrast is a separate problem. Solder, copper and laminate attenuate X-rays differently, and the difference is small for materials of similar density. Where a void sits against a background of similar absorption, it can be invisible in the reconstruction even though it is physically large. Longer exposure improves the statistics but lengthens the scan, so contrast decisions are always a compromise. Our comparison of X-ray and AOI explains how the two methods divide the work.

Void Analysis in Solder Joints

Void measurement is one of the most common uses of CT. A two dimensional image shows a void as a light region but cannot say whether it is a single sphere or several connected cavities, or how far it sits from the interface. A reconstructed volume answers all three questions.

Void percentage is calculated by comparing void volume with total solder volume, which is a more meaningful figure than the area fraction taken from a shadow image. The distribution matters as much as the total: voids concentrated at the interface between solder and pad affect reliability far more than the same volume dispersed through the bulk. Position therefore belongs in the report alongside the percentage.

BGA and Package Inspection

For area array packages, CT shows every ball in the array without the superposition that limits two dimensional inspection. Cracked balls, head in pillow defects and mid array bridging become visible, and the shape of each ball can be compared against its neighbours to identify systematic process problems.

Package on package assemblies are the hardest case, because the upper and lower ball arrays overlap in projection. Reconstruction separates them, which is often the only way to confirm that both rows of joints are sound without destructive analysis. The same applies to stacked die packages and to assemblies where a heat spreader covers the joints of interest.

Plated Hole and Barrel Analysis

Plated through holes are another strong application. A barrel crack that runs partly around a hole may not appear as an open circuit, and a two dimensional image cannot distinguish a crack from a thin plating section. CT shows copper thickness, barrel cracks and the condition of the knee where plating meets the surface layer.

This is used both for failure analysis and for process qualification, particularly on thick boards where plating uniformity is harder to achieve. Our notes on hole copper explain why thickness distribution varies with aspect ratio and current density.

Sample Preparation for Failure Analysis

CT is non destructive, which is its main advantage in failure analysis: the sample remains intact for further testing. Boards should be cleaned and, where possible, cut to the region of interest so that magnification can be increased. Loose material must be removed, because a fragment that moves during the scan produces artefacts. Where the sample is potted to hold it steady, the potting material should be chosen for low X-ray absorption.

The scan should be planned around a specific question. Scanning an entire board at low resolution produces a large dataset and few answers, while scanning a single component at high resolution usually identifies the mechanism directly. Deciding the question first saves both time and reconstruction effort, and it keeps the dataset small enough to review properly when the analysis has to fit around a production schedule.

Interpreting Results Without Over-Reading

Reconstruction artefacts are the main risk. Beam hardening, ring artefacts and scatter can all create features that look like defects, and a reviewer who is not familiar with them will find voids that do not exist. Comparing the suspect region with a known good sample is the simplest guard against this.

Quantitative results should also be treated with care. Void percentage from a CT scan depends on the threshold used to separate void from solder, and small changes in that threshold shift the number. Reporting the threshold alongside the result makes the measurement reproducible.

Where CT Fits in the Quality Toolbox

CT is slow compared with optical or conventional X-ray inspection, so it is not a production tool. It belongs in failure analysis, in process qualification and in the investigation of a specific problem that other methods have narrowed down. Used that way it is extremely efficient, while treating CT as a screening tool produces large volumes of data and very few decisions.

The sequence matters. Optical and X-ray inspection identify the suspect location, CT characterises the defect in three dimensions, and destructive cross sectioning confirms the metallurgy if it is still unclear. Root cause feeds back into the process, which is how the loop described in our production process flow notes is meant to work.

FAQ

What is the difference between X-ray inspection and computed tomography? Conventional X-ray inspection produces a two dimensional projection in which all layers overlap, while computed tomography reconstructs a three dimensional volume from many angles. CT resolves hidden joints and internal structure that a projection cannot separate.

Can CT measure void percentage in a BGA? Yes, and it is more accurate than an area measurement from a shadow image because it uses volume. The figure still depends on the threshold used to define a void, so the threshold should be stated with the result.

Is CT fast enough for production inspection? No. A scan takes far longer than optical or conventional X-ray inspection, so CT is used for failure analysis, process qualification and problem investigation rather than for routine screening of every board.