Parts With Different Heights: Depth of Field, Occlusion and Focus

When a stepped part is imaged from above and one level looks soft, the first instinct is usually to change the lens. In most cases the fault is one of two different things, and they need different corrections. If the softness covers the whole level uniformly, the depth of field is too small for the height difference, and focus position or aperture can fix it. If a band of the lower surface cannot be seen at all while everything else is sharp, the cause is geometric occlusion by the step wall — and no focus or lighting setting will recover it.

Separating focus problems from occlusion

Depth of field is the range of object distances around the focused plane that still resolves acceptably. Object-side, it follows DOF ≈ 2 · F# · CoC · (1 + M) / M². At M = 0.042, F# = 4 and CoC = 0.015 mm this is roughly 70 mm, so a few millimetres of step height is not the limiting factor in a large-field station. Occlusion behaves differently: the hidden width is w ≈ h · tan α, where h is the step height and α is the chief-ray field angle at that point, α ≈ arctan(r / L). With h = 5 mm, r = 85 mm and L = 200 mm, α ≈ 23° and w ≈ 2.1 mm — while the same step at the centre of the field is barely occluded at all.

One image separates the two. If the unreadable region is narrow near the centre of the field and widens towards the edges, the cause is occlusion. If sharpness falls off uniformly across the whole level instead, the cause is depth of field or focus.

The numbers to check before the station is built

• Measure the real step height across the tolerance range, not the nominal drawing value.
• Require h ≤ 0.7 × DOF with focus at mid-height — the remaining 30 % covers focus repeatability, part flatness, fixture variation and temperature.
• For a fixed-focal lens, expect a scale error of roughly h / WD between the two levels: at a 300 mm working distance a 5 mm step is about 1.7 % of scale, which is systematic and does not average out.
• Confirm the occlusion band on a sample part with a scale target beside the wall, and compare it with h · tan α.

When the answer is optics rather than settings

Focusing at mid-height shares the available depth of field between the two levels instead of spending it on one. Stopping down adds margin in proportion to the f-number, at the cost of light (1 / F#²) and, eventually, diffraction. Neither touches the occlusion band. Removing it requires removing the chief-ray tilt: an object-side telecentric lens keeps α ≈ 0 across the field, so the band disappears and both levels share a single scale factor. Where the height spread exceeds any achievable depth of field, the alternatives are a stored focus position per recipe, an electrically focus-tunable lens, or an extended-depth-of-field image stack.

Related reading: building a telecentric measurement station, telecentric versus standard FA lens compared, and the telecentric lens range. The detailed questions on this topic are answered below — to have the numbers run against your own part, send us the drawing and the tolerance.