The simplest demonstration of parallax requires no equipment. Hold a finger up in front of your face, close one eye, then switch eyes. The finger appears to jump sideways, because each eye views it from a slightly different position.
Why X-ray systems have parallax at all
X-ray photons are generated in an extremely small region of the target inside the source, which makes it effectively a point source. Photons leave that point as a diverging cone, pass through the sample, and cast a shadow onto the detector.
Because the beam diverges, only the ray traveling straight down the center axis passes through the sample perpendicular to it. Every other ray arrives at an angle, and the further a feature sits from the central axis, the more oblique the angle through which it is viewed.
A feature at the center of the field is imaged nearly straight through. The same feature near the edge of the field is imaged from the side.
Parallax as a distortion
This produces a real geometric distortion. Even a perfectly uniform sample will not look uniform across the image, because the edges are being viewed at a different angle from the center.
It also means the apparent lateral position of a feature depends on its height within the sample. Two features stacked vertically, one above the other, do not project onto the same point of the detector unless they sit directly on the central axis. The higher one shifts relative to the lower one, and the size of that shift depends on both the height difference and the distance from the axis.
Parallax as a feature
That shift is exactly what makes oblique inspection work.
In 2D, 2.5D, and side-view inspection, those angled rays are the point. They provide a perspective view of the sample rather than a flat overhead projection, which is what allows structures at different depths to be seen as separate rather than superimposed.
BGA inspection is the clearest application. A solder ball has two interfaces that matter: the joint to the PCB at the bottom and the joint to the component at the top. In a straight-through view they sit on top of each other and cannot be evaluated independently. Viewed obliquely, they separate, precisely because features at different heights shift by different amounts. The distortion is what delivers the information.
Parallax as a problem
Computed tomography inverts the situation.
CT collects many 2D projections of the same object and reconstructs them into a volume, and reconstruction depends on knowing exactly where every feature was relative to the geometry in every projection. Parallax has to be characterized and corrected rather than exploited, or it introduces error into the reconstruction.
The good news is that collecting many angles is what makes correction possible. Multiple views of the same object from known positions provide enough information to determine what parallax is doing and account for it in the algorithm.
Why limited-angle acquisition cannot fully reconstruct
This is where the two halves connect.
Planar CT and other limited-angle techniques collect projections across a restricted range of angles rather than all the way around the sample. That restriction is usually practical rather than optional, since a large populated circuit board cannot be rotated freely inside an inspection system.
The consequence is that a limited-angle acquisition does not capture enough angular diversity to constrain a full volumetric reconstruction. There is not enough parallax information available for the algorithm to resolve the object completely in three dimensions. What it can produce instead are layer images parallel to the sample plane, which is why laminography reconstructs slices through a joint rather than a freely sliceable volume.
That limitation is not a defect in the technique. It is a direct consequence of the geometry available, and understanding it explains why electronics inspection settled on the tools it did.
Both at once
Parallax is a distortion in X-ray imaging, and it is also one of the more useful properties the geometry provides. Whether it is a problem or a feature depends entirely on what the inspection is trying to accomplish, and knowing which situation applies is a large part of getting a useful image.
