Fault coverage vs. throughput in x-ray inspection

Peter Edelstein

Q: Where should x-ray inspection systems concentrate their efforts?

A: There is no simple answer to that question. X-ray inspection can identify solder-joint defects that no other test method (such as automated optical inspection or in-circuit test) can find. But you have to trade off fault coverage, false-call rates, and throughput. Some x-ray approaches are time-consuming, lending themselves primarily to high-value situations with low volumes or to product sampling for process modifications. We also contend that some methods are mechanically complex enough to make setup and adjustment more difficult, which reduces their ability to handle high-mix situations. Image resolution and analysis techniques determine what a particular method can find and the likelihood of false failures.

Q: Are you talking only about automated systems?

A: Not necessarily. Manual x-ray systems can provide very high-quality images, allowing close examination of a board’s solder joints and therefore its faults. Such systems are generally less expensive than the automated variety, but they are also much slower. They usually use 2-D transmission technology, and they can experience false failure rates as high as 10,000 false failures per million solder joints. They work most effectively during product development or for product sampling or very low-volume production.

Q: What factors determine which is the best solution?

A: The answer depends on who the customer is and what kinds of products the customer produces. In both automotive and medical applications, for example, customers suffer catastrophic consequences in the event of system failure. Medical devices are generally high-value products that can absorb additional costs to insure against faults. Car makers run at high volume on relatively thin margins and will not tolerate any added costs. They can’t afford faulty products to escape, but for them, false failures are of particular concern. Cell-phone manufacturers also run at high volumes and low margins, but failures are more annoying than dangerous.

Q: Is 3-D imaging necessary?

A: We believe it is, although perhaps not at every location on every board. There are several ways to achieve it.

Laminography techniques produce “slices” of the board at different depths. They require a moving image train, which makes the systems mechanically more complicated.

Our alternative—off-center tomosynthesis—gathers image information using a static image train and then applies software algorithms to digitally construct a 3-D representation of the object under inspection. Using this approach, we can provide 2-D and 3-D information from the same data.

We have obtained excellent fault-coverage results and significantly reduced false-failure rates with this approach. Laminography is still common, but we expect the popularity of tomosynthesis to grow rapidly as new machines adopt the technique.