Glass and Transparent Part Inspection: Solving the Ghost Edge Problem

Why a transparent part shows more than one edge

Light entering a glass part refracts at the first surface and again at the second. Each refraction produces an apparent edge in the image, and the two do not coincide. For a parallel plate viewed straight on, the displacement is small; as soon as the part is tilted or has a curved surface, the two edges separate visibly. A third, fainter edge often appears from light that reflects internally before exiting. The inspection therefore sees a stack of edges, only one of which corresponds to the dimension being controlled.

Identifying which edge is the real one

Change the illumination from bright field to dark field, or move a backlight to a low-angle position. The true geometric edge stays in place; refraction-generated edges shift or vanish because their position depends on the light path. This test is more reliable than judging by edge sharpness, because on clean glass the ghost is often sharper than the true edge.

Why telecentric viewing helps on transparent parts

A telecentric lens accepts only rays nearly parallel to the optical axis. Refracted rays leave the glass at an angle, so most of them fall outside the acceptance cone and never reach the sensor. The practical effect is that ghost edges weaken markedly compared with a conventional lens, even though the true edge is unaffected. This is the main reason telecentric imaging is specified for glass and transparent part measurement.

Choosing the configuration

Across the POMEAS telecentric range, magnification spans 0.061X to 12.5X and working distance spans 9.5 mm to 468 mm. Transparent parts usually need a longer working distance so that the backlight can be placed far enough away that its angular spread at the part is small. Telecentricity, quoted at 0.02 to 0.04 degrees on the models where it is specified, directly determines how strongly off-axis refracted rays are rejected.

What the tolerance should be written against

State whether the controlled dimension is the outer envelope, the wall thickness, or a feature on one surface. Each maps to a different edge in the image, and if the specification does not say which, two competent inspectors will measure two different edges. This is one of the few cases where the measurement disagreement is a specification problem rather than an equipment problem.

When to move away from pure backlight

If the part is thick relative to the feature being measured, refraction displacement can exceed the feature itself and backlight stops being usable. In that case a coaxial or dark-field arrangement that images only the top surface, combined with a telecentric lens, gives a cleaner single edge. The trade-off is that surface condition then matters, so cleaning and handling become part of the measurement process.

Frequently asked questions

Why do I see two edges on a glass part?

Light refracts at both surfaces, producing two apparent edges. A third can appear from internal reflection.

How do I know which edge is real?

Change the illumination angle. The true geometric edge stays fixed while refraction edges shift or disappear.

Does a telecentric lens remove ghost edges on glass?

It weakens them substantially, because refracted rays leave at an angle and fall outside the narrow acceptance cone.

What working distance suits transparent parts?

A longer one, so the backlight can sit far enough away that its angular spread at the part is small. The POMEAS range spans 9.5 mm to 468 mm.

Why do two inspectors disagree on a glass measurement?

Because the specification does not say which edge is controlled. Outer envelope, wall thickness and surface features map to different edges.

When is backlight unsuitable for glass?

When refraction displacement exceeds the feature being measured, for example on thick parts with small features.