Telecentric illumination: measuring round shiny parts accurately

Shiny round parts are the worst case for backlit gauging. A diffuse backlight wraps light around the curved edge of a polished shaft or O-ring, the edge reads bright where it should be black, and the diameter comes out small — by an amount that moves with surface finish. Telecentric illumination fixes this: a collimated light behind the part, matched to a telecentric lens, sends only parallel rays into the optics, so the silhouette edge lands where the metal actually is. This explainer covers the mechanism, the 20-30% depth-of-field gain that comes with it, a fully worked example on a Ø6 mm pin at ±25 µm, and when a plain diffuse panel is still the right buy. Shafts, tubes, screws, springs and O-rings are the classic applications.

What telecentric illumination is

A telecentric illuminator is a collimated light — every ray leaves the unit parallel to the optical axis within a small cone — mounted behind the part and aimed at a telecentric lens. The lens does the other half of the job: its chief rays stay parallel to the optical axis, with the entrance pupil effectively at infinity, so magnification does not change when the part shifts along the axis. Break the pair and you fix one error while keeping the other. When you need a telecentric lens for gauging sets out the lens half of that decision, and entocentric vs telecentric lenses the trade against ordinary glass.

In the shop these units sit in the collimated lights category, filtered by beam size, colour and voltage, and several series are built to mate with a named telecentric lens family — worth using when your lens belongs to one, because beam diameter and numerical aperture then line up by design.

The border effect: why shiny cylinders measure small

A diffuse panel radiates at every angle. At the curved shoulder of a polished shaft, wide-angle rays strike the surface just inside the true edge and mirror into the lens: the camera sees bright background where metal is, and the edge detector draws the boundary inside the part. The measured diameter reads low, commonly by a few pixels on ground steel, and the bias moves with polish, curvature and part orientation — which is why it resists calibration.

Collimated rays behave differently. A ray either clears the tangent line of the cylinder and travels straight into the lens, or it hits the curved shoulder and reflects away at a steep angle outside the lens acceptance cone. The transition happens at the tangent line itself, so the detected edge coincides with the physical edge. This "border effect" is what telecentric illumination removes, and it is the largest correctable error in backlit gauging of polished round parts. The broader collimated-versus-diffuse choice for edge work is covered in collimated backlight or diffuse backlight for edge gauging, and the reflection law behind it in bright field vs dark field lighting. If readings drift shift-to-shift on parts that measure fine on a micrometer, put edge reflection on the suspect list from micron-level image measurement accuracy.

Numerical aperture: where the depth-of-field bonus comes from

The illuminator's numerical aperture (NA) is lower than the lens's. Only rays the light actually sends can form the image, so for depth of field the system behaves as if the lens had the illuminator's NA, while resolution stays set by the lens's own NA. On real gauging systems the gain is about 20-30% more usable depth of field; the three levers that move it are lens type, wavelength and pixel size.

It is the same physics as stopping down a conventional lens: a smaller aperture passes more collimated rays and the field deepens. In a cell, that means part-height variation and fixture tolerance stop eating your edge sharpness — how to prove it is in telecentric depth of field: validate measurement accuracy.

A second consequence is standoff. Because coupling between light and lens is efficient, the part-to-light distance can grow without the image falling apart — useful when a chute, guard or fixture has to sit between them.

Throughput: short exposures on moving parts

Concentrating the flux into a narrow cone puts more of it inside the lens, so exposures drop. Blur equals speed times exposure: at 100 mm/s, every 10 µs of exposure smears the edge by 1 µm. Short strobes hold that under a pixel, and strobing instead of running continuous also stretches LED life, a trade quantified in LED lifetime L70: why strobe mode makes lights last longer. High-speed sorting lines are the classic beneficiary: short pulse, sharp silhouette, part gone.

Worked example: a Ø6 mm polished pin at ±25 µm

Inputs. Diameter 6 mm ground dowel pin, drawing tolerance ±25 µm. Accuracy target from the 10-to-1 rule in vision metrology: the 10 to 1 rule: ±2.5 µm. Sensor: a common 2/3-inch metrology camera with 3.45 µm pixels and 8.8 mm sensor width (example values — swap in your camera's). Handling: V-block centring, part passing at 100 mm/s, strobed on the fly.

Parts. The lens is the LS-TCL-1972, a 1x telecentric lens with a 440 mm working distance, 2/3" sensor format and C-mount. The light is the LS-CLM-0051, a 35 mm collimated spot light.

Arithmetic. At 1x the field of view equals the sensor width: 8.8 mm. The 6 mm pin fills 68% of it (6 / 8.8 = 0.68), leaving 1.4 mm per side for centring. Object-space pixel size is 3.45 µm / 1 = 3.45 µm — the same conversion used to calculate object space pixel size from magnification. A tenth-of-a-pixel edge fit locates each edge to 0.345 µm; a diameter needs two edges, so √2 × 0.345 = 0.49 µm of repeatability, one fifth of the ±2.5 µm target, leaving budget for calibration and lens errors.

Now the light. On a diffuse panel, call the border effect two pixels lost at each edge (measure yours with the tip above): 4 × 3.45 µm = 13.8 µm of diameter error, 55% of the tolerance, moving with surface finish. With the collimated spot that term is zero, and the pixel grid plus subpixel fitting carries the measurement.

Motion: at 100 mm/s a 20 µs strobe smears the edge by 100 × 0.00002 = 2 µm, or 0.58 px — negligible. The same scene on a diffuse panel needing 200 µs for equal grey level smears by 20 µm, nearly 6 px, and the edge detector chases a ramp instead of a step.

Beam check: a 35 mm beam against an 8.8 mm view gives four-times coverage, so light alignment is forgiving and the unit can sit back clear of the fixture.

Decision. Order the 1x telecentric lens with the 35 mm collimated spot and a strobe channel on the controller. The pairing holds edge repeatability near 0.5 µm against a ±25 µm tolerance; on the diffuse panel the border effect alone consumes over half the tolerance before anything else goes wrong.

Choosing the beam, the lens and the colour

Size the beam to the view, not the part: beam diameter at least 1.3× the field of view, so every position the part can occupy stays inside the parallel-ray zone. Round telecentric illuminators run from 35 mm spots to the 200-400 mm class; a shaft wider than the largest beam in the collimated lights finder goes back to a diffuse panel with a calibrated offset.

Pick the lens first, then a light that mates with it, since several collimated series are named for the lens family they feed. Remember the light sits behind the part and its body length adds to cell depth — 223.9 mm on the 90 mm beam class.

Colour: red with a monochrome camera is the default, and the reasoning is in light color and monochrome camera contrast.

The workhorse shape for shaft and tube cells:

Compact spots for small fields of view:

Lens and light can be pulled together in the telecentric lens finder and the light finder, or sized in one pass with the system configurator.

Flat collimated panel or round telecentric illuminator?

Two collimated routes exist. The LS-BKL-0830 is a 120 x 120 mm collimated back light; the LS-CLM-0079 is a telecentric illuminator with a 90 mm beam.

Take the flat panel when the part is wide, the cell is shallow, or the tolerance sits at the ±0.1 mm class. Take the matched round illuminator when the tolerance is in tens of micrometres on a polished, curved edge — that is where its NA matching pays.

When the usual answer is wrong

• Matte parts. A blasted or black-oxide shaft scatters light at the edge instead of mirroring it; a diffuse panel already reads it true, and the collimated spend buys nothing.
• Transparent parts. A clear rod deviates rays by refraction, not reflection, and its shadow can read larger than the rod. Calibrate against a certified master or change technique.
• Wide parts. Beyond the 200-400 mm beam class, collimation is out of catalogue reach; a large diffuse panel with a calibrated offset is the practical answer.
• Entocentric lens. The light alone cannot stop magnification changing with part height; gauging needs the lens too (the FAQ quantifies it).
• Loose tolerances. At ±0.5 mm on a matte part, spend the money on fixturing, not photons.

For where telecentric lighting sits among the other geometries — bright field, dark field, dome, coaxial — read machine vision illumination types.

Checks before you order

• Beam diameter at least 1.3× the field of view, and the light body length fits the cell depth.
• Lens magnification × sensor width covers the part plus centring tolerance; object-space pixel no bigger than your accuracy target, with tenth-pixel edge fitting carrying the rest.
• Lens working distance clears the fixture, with the light behind the part and the beam covering the view.
• Light colour matched to the camera: red or blue with a monochrome sensor.
• A controller channel that strobes, triggered by encoder or photocell.
• Prove it with 30 repeats on a certified pin and a micrometer comparison before the cell goes live.

Next step: run the system configurator from the light step, or bring the part and its tolerance to us through the contact page and we will size the pair with you.