Automated X-Ray Inspection For Hidden Solder Joints

Optical inspection can only see what is visible. A ball grid array, a land grid array, a quad flat no lead package, and the underside of a connector all hide their solder joints behind or beneath the package, and the joints that are hardest to see are often the ones carrying the most current or the highest stress. X-ray inspection looks through the assembly instead of at it, and the image it produces shows the solder by the difference in how much radiation it absorbs compared with the board and the package.

This article explains how the image is formed, what a two dimensional and a three dimensional system each show, what can and cannot be judged, and how limits are set so that the inspection is useful.

How The Image Is Formed

A source emits radiation that passes through the assembly and reaches a detector on the other side. Dense materials absorb more than light ones, so a solder joint appears dark against the lighter laminate and the package, and a void inside the solder appears light because there is less material in the path. The contrast in the image is therefore a map of the material thickness along each ray, which is both the strength and the weakness of the method: it shows the solder, and it also shows everything else in the same line of sight.

The resolution depends on the size of the source spot, on the geometry, and on the detector, and the magnification is obtained by moving the board closer to the source. A smaller spot gives a sharper image but less power, and a thicker or denser board requires more power to penetrate, which forces a larger spot and a softer image. The tube voltage and current are therefore set for the assembly rather than chosen once, and a board with heavy copper and a large metal stiffener is much harder to image than a thin board.

Two Dimensional And Three Dimensional Systems

A single projection of a ball grid array shows the top row of balls, the bottom row of balls, and the pads on both sides, all superimposed. That is adequate for a void measurement when the ball is the only dense object in the path, and inadequate for anything more subtle, because a defect in one row can be hidden behind a good joint in another. Oblique views, taken at an angle, separate some of the rows and are the traditional way of looking at the joint at the edge of a package.

A computed tomography or laminography system reconstructs the assembly in three dimensions, which allows a single layer to be examined without the interference of the others. That makes a void measurement more accurate, allows a head in pillow defect to be seen at the interface between the ball and the pad, and makes the inspection far less sensitive to the orientation of the board. The cost is time: a reconstruction takes longer than a projection, so the technique is applied to a sample or to a critical package rather than to every joint on every board.

What Can And Cannot Be Judged

A void is the classic measurement, because it is a difference in density and appears directly in the image. The result is expressed as a percentage of the ball area at its widest point, and the criterion distinguishes between one large void and a scattering of small ones, since a single void that spans most of the interface is more serious than the same total area distributed. An open joint, a bridged joint, and the fill of a plated hole are all visible with a suitable setup, and the fillet of a package with exposed joints can be judged from the side.

What the method cannot see is the quality of the interface. A joint that is metallurgically poor but geometrically sound appears the same as a good one, because both contain the same amount of solder in the same shape. A crack through the bulk of a joint is visible only when it is large enough to change the density along the path, and a thin crack at the pad interface is effectively invisible. X-ray is therefore a geometric measurement rather than a metallurgical one, and the distinction matters when it is used to accept a process.

Setting Limits And Avoiding False Calls

The limits come from a standard, from the customer, or from a reliability study, and they are expressed in a way that the machine can measure. A void criterion of a quarter of the ball area is a common starting point, but the figure has to be tied to the measurement method, because the same joint measured on two systems with different thresholds produces different numbers. The machine settings, the reconstruction parameters, and the definition of the ball boundary all belong in the specification.

The trade between sensitivity and false calls is the same as in optical inspection. A program that flags every small void sends good product to rework, and a program that is relaxed lets a real defect pass. The practical control is to correlate the X-ray result with a destructive section on a sample, which shows what the machine is actually measuring, and to adjust the limits from that correlation rather than from the general rule.

Program, Throughput And Design

The inspection program is built from images of the product, and the inspection time rises with the number of views and with the reconstruction. Most lines therefore inspect the critical packages at full resolution and the rest at a lower level, or they inspect a sample of panels rather than every unit. The decision is based on the consequence of an escape, which is high for a large ball grid array on a safety related product and low for a package whose joints are also visible optically.

The layout affects the result. Overlapping packages, a large metal shield, or a board with heavy copper makes the image harder to interpret, and a design that keeps the dense packages apart helps. A test coupon that carries the same ball pattern in an accessible area gives a reference point, and it is worth including where the process is being qualified. The assembly defects that the inspection is looking for are described under component shift during reflow, the acceptance framework under PCB design quality characteristics, and the layout consequences under layout decisions that affect production.

Additional Considerations for This Build

Practical attention to BGA pays for itself here, because it decides whether the finished board behaves as the drawing intended. Where the requirement is not stated on the fabrication drawing or in the assembly notes, the shop has to assume a default, and that default is rarely the value the design was simulated with. Stating BGA explicitly, together with the tolerance that applies, removes the assumption and keeps the result predictable from batch to batch.

FAQ

Can X-ray see a cracked solder joint? Only when the crack changes the amount of material along the beam. A thin crack at the pad interface is effectively invisible, which is why the method is a geometric check rather than a metallurgical one.

Is a two dimensional system enough? For a void measurement on a single row of joints, often yes. For an interface defect inside a ball or for a package with several rows, a three dimensional reconstruction is far more reliable.

Why do two machines give different void percentages? Because the boundary of the ball and the threshold for what counts as a void are defined by the program and by the reconstruction settings. The figures are comparable only when the method is the same.