Laminography: 3D X-Ray Imaging for Solder Joints
Laminography, also called computed laminography or tomosynthesis, produces an X-ray image of one plane inside an assembly while the material above and below that plane is blurred out. The result is a view of a single solder joint layer without the superposition that makes a conventional two dimensional image hard to read.
The technique sits between ordinary transmission X-ray, which is fast but flat, and full computed tomography, which is precise but slow and often blocked by a large board. It is the practical answer for assemblies where several rows of joints overlap in the beam direction.
What Laminography Adds
A two dimensional X-ray image is a projection of everything the beam passes through. On a package with two or three rows of balls, the image of the front row is overlain by the rows behind it, and a void in one row can be confused with a feature in another.
Laminography removes that ambiguity for the plane of interest. Rows behind and in front of the chosen plane are blurred into a background, so the operator can judge the shape of a fillet, the position of a crack and the size of a void on the plane that matters.
How the Image Is Built
The system captures a series of projections while the source and detector move relative to the board in a defined geometry. The projections are then reconstructed mathematically, with the reconstruction tuned so that only one depth remains in focus.
Two geometries are common. In one, the source moves in a linear path and the detector moves in the opposite direction; in the other, the motion is circular or elliptical. The choice affects which directions remain blurred and therefore how well the technique performs on a given board layout. A linear scan blurs along one axis and acquires quickly, while a circular scan blurs in every direction within its plane of travel and copes better with joints surrounded by dense copper. The trade is acquisition time, which rises with the number of projections.
Resolution and Slice Thickness
The depth of the in focus region is set by the geometry and by the number of projections. A small slice thickness resolves fine detail but takes longer to acquire and may not cover the full height of a solder joint in one image.
Typical systems resolve features of the order of a few micrometres in plane, with slice thicknesses from about 10 um up to 100 um depending on the setting and the magnification. The slice thickness should be chosen to match the question being asked: a void measurement needs a thin slice, whereas a quick look at fillet shape tolerates a thicker one.
Inspecting Solder Joints and Voiding
For a ball grid array, the useful planes are the package interface, the mid ball and the board interface. Voiding is normally assessed at the plane where the void is largest, which for a reflowed ball is usually near the board side interface where the solder has flowed around the pad.
Void area is expressed as a percentage of the joint cross section, and the acceptance limit comes from the applicable standard for the product class. A measurement taken on a thick slice overestimates the void because the in focus region includes more of the spherical void than a thin slice would.
Through-Hole and Press-Fit Applications
The same technique is useful on plated through holes and press-fit pins, where the interest is the distribution of copper along the barrel and the presence of voids in the plating. Slices taken along the axis of the hole show the copper thickness at the surface, the centre and the knee.
This makes laminography a useful screening tool before committing a board to the microsection saw. The copper in the hole can be checked quickly on several holes, and the section is then reserved for the hole that the scan suggests is worst. On a press-fit connector in a thick backplane, the scan also shows whether the pin has deformed the barrel, a distinction that a two dimensional image usually cannot make.
Limits: Dense Boards and Shadowing
Dense copper and thick metal parts block the beam. A board with heavy ground planes, a metal stiffener or a large heat sink will scatter radiation and reduce contrast, and in the worst case the plane of interest is simply not visible.
Assembly geometry adds a second limit. Tall components near the package can shadow the region of interest depending on the beam angle, so a joint that is easy to image in one orientation may be unreadable in another. The inspection program should therefore be built around the real assembly rather than an idealised layout.
Recipe Setup and Programming
A recipe defines the region of interest, the plane height, the geometry, the number of projections and the analysis thresholds. Setting it well takes time on the first article, and that time is repaid every time the job runs again.
The plane height should be set from a physical reference, such as the top of the board surface measured with the system height sensor, rather than from a nominal value taken from the CAD model. Board thickness tolerance and warpage both move the plane, and a plane set nominally will drift off the interface across a large panel.
Throughput, Sampling and Cost
A laminographic scan is slower than a two dimensional image and much faster than full tomography, so it is normally used as a second level check rather than on every board. A common arrangement is a fast two dimensional pass that flags suspicious joints, followed by laminography of those areas.
Machine time should be spent where the risk is concentrated. Corners and centre joints of a large package, joints near a heavy thermal pad and joints on the worst warpage axis return more information than a random sample, and a small control set from a known good build keeps the interpretation honest.
Correlating Results With Cross Sections
The scan is an indirect measurement, so its interpretation has to be calibrated against something physical. Sectioning a joint that the scan identified as defective, and one it identified as good, confirms that the criteria in the recipe mean what the operator thinks they mean.
The microsection preparation has to be precise for this comparison to be fair, because a section taken off centre will show a smaller joint than the scan did. Once correlated, the scan can be trusted on its own for routine work, and the section reserved for confirmation and for failure analysis.
FAQ
What is the difference between laminography and computed tomography? Both reconstruct a three dimensional view from many projections. Laminography keeps a single plane in focus with a limited motion, while tomography reconstructs the entire volume, which takes longer and needs access from many angles.
Can laminography measure void percentage in a solder joint? Yes, when the slice thickness is small enough and the joint is fully within the field of view. The slice setting changes the apparent void size, so the measurement is only comparable when the recipe is held constant.
Does laminography replace cross sectioning? No. It is a non destructive screening method that identifies which joints deserve a section. The section remains the reference measurement for structure and thickness.
}
