
Both techniques reconstruct three-dimensional information from a series of X-ray images. They arrive there by completely different routes, and the reason two approaches exist comes down to the shape of the thing being inspected.
Start with the shadow
Every X-ray image begins the same way. A source emits photons, a sample sits in the beam, and a detector records what makes it through.
The result is a density map. Where the sample is thin or less dense, more photons arrive and the image is brighter. Where it is thick or dense, fewer arrive. That shadow is the raw material for everything that follows.
Magnification comes from position. Moving the sample toward the source magnifies it, because the diverging beam projects a larger shadow. Moving it toward the detector shrinks the image and widens the field of view.
Both are useful, for different jobs. Inspecting an individual solder ball on a BGA calls for maximum magnification, so the sample goes as close to the source as possible. Counting components across a full 17-inch reel calls for the opposite, so everything sits in frame at once and can be counted quickly.
How conical CT works
Conical CT is the technique behind a hospital CAT scan. In a medical scanner, a source and a ring of detectors rotate around a stationary patient at speed. In an industrial system the arrangement is usually reversed: source and detector stay fixed while the sample rotates a full 360 degrees between them.
Either way, the principle is complete angular coverage. Projections are collected from all the way around the object, and a reconstruction algorithm assembles them into a volume.
For a compact object this works extremely well. At every rotation angle the detector receives a meaningful projection, and the algorithm has full information to work with.
Why conical CT struggles with flat objects
Rotating a printed circuit board through that same 360 degrees exposes two problems, and both are geometric rather than fixable in software.
The board disappears edge-on. Through part of the rotation, the beam passes along the plane of the board rather than through it. The projection collapses to a line, and at those angles the detector receives essentially no usable information about the sample’s interior. A meaningful fraction of the rotation contributes nothing.
Rotation forces the sample away from the source. A board has to clear the source and detector throughout its full rotation, which means it must be positioned far enough back to swing freely. Distance from the source is precisely what limits magnification, so the requirement to rotate directly costs resolution on the small features that motivated using CT in the first place.
Those two problems compound. The technique loses information at some angles and loses resolution at all of them, on exactly the class of sample most electronics inspection deals with.
How planar CT works
Planar CT, named for the plane it works within, solves both problems by not rotating the sample.
Instead the source and detector move in a coordinated way to collect images of the sample from a range of oblique angles. Some implementations move the source and detector while the sample stays fixed. Others hold the source and detector together and move the sample. The objective is identical: acquire a set of angled views without ever putting the sample edge-on to the beam.
Sixteen, thirty-two, or sixty-four images are typically enough. Because the sample never has to swing through a full rotation, it can sit very close to the source, which preserves the magnification that conical CT gives up.
What limited angle costs and buys
Collecting sixty-four images across a restricted angular range obviously provides less information than collecting thousands across a full rotation. This is limited angle, or partial, CT, and the reduction is real.
What it buys is speed, and the difference is not incremental. A planar CT acquisition completes in a couple of seconds, which is what makes it viable inside an SMT line where a conical CT scan measured in minutes simply cannot fit.
It also turns out that the full dataset is unnecessary for the question being asked. Determining whether a BGA joint has excessive voiding, or evaluating a QFN pad, does not require thousands of slices. Somewhere around 100 to 128 slices is sufficient to make that determination reliably. Collecting far more data would produce a more complete volume and would not produce a better answer to the question actually being asked.
That is why planar CT became the standard for SMT inspection.
Different hardware, different software
The two techniques require genuinely different machines. The motion systems are built around different requirements, and neither arrangement produces the other by adjustment.
The software differs as much as the hardware. Planar CT software is oriented toward manufacturing, with built-in tests for BGA inspection, void measurement, and the specific pass-fail determinations a production line needs. Conical CT software is aimed at analysis, with tools for porosity, crack detection, and the kind of investigation that happens when someone is trying to understand a failure rather than screen for one.
Which to use
For flat samples, which means most electronics, planar CT is the right tool. For objects that are not flat, a connector body, a machined part, or a pin being examined for internal cracks, conical CT is the one that answers the question.
Because both cases arise in most operations eventually, systems offering both modalities on a single platform are worth seeking out, as is the ability to add the second modality later. It is a considerably easier problem to solve at purchase than after the fact.