Telecentric Lens for Dimensional Measurement and Inspection

Dimensional measurement with a telecentric lens works because the lens, not the software, removes the two errors that ruin optical gauging: perspective magnification drift and field distortion. A telecentric lens holds magnification constant as the part moves in Z and keeps distortion at 0.02–0.1%, which turns a camera into a non-contact measuring instrument at the micron level. The specification work is in matching magnification, sensor and illumination to the tolerance.

This article is about building a measurement station rather than choosing a single component. It covers what the optics must deliver, how to convert a part tolerance into a magnification and field of view, how illumination determines whether an edge can be found at all, and how the same principles play out in 3C electronics, battery and precision hardware inspection.

What dimensional measurement demands from the optics

A vision system used for detection only has to deliver a recognisable image. A vision system used for measurement has to deliver a scalable image — one where a fixed number of pixels corresponds to a fixed physical distance, regardless of where the part sits or which part is loaded. Four optical properties carry that requirement:

The last row is the one most often skipped. A lens that resolves 16.78 µm on a 10 MP sensor wastes the sensor, and a lens that out-resolves a 2 MP sensor wastes the lens. Resolution is a matched pair.

The measurement chain, and where accuracy is actually lost

Optics are the first link, not the only one. Accuracy on the line is the sum of contributions from every stage, and the largest term is usually not the lens.

Note what a telecentric lens does and does not remove. It removes the Z-position term from the optics and removes distortion from the scale. It does not improve edge detection under poor illumination, and it does not compensate for a fixture that rotates the part. Those are separate engineering tasks.

Converting a part tolerance into optics specifications

The design sequence runs backwards from the tolerance, not forwards from a lens catalogue.

1. State the tolerance and the decision. A ±20 µm tolerance with an accept/reject output allows roughly a third of that — about 7 µm — to be spent on optics, sensor and illumination combined.
2. Size the field of view. Take the workpiece bounding box and add margin. The existing selection guide recommends against setting the field of view exactly equal to the part size, because placement tolerance will push features off the sensor. Add 10–20%.
3. Compute magnification. Magnification = effective sensor dimension ÷ required field of view. With a 2/3″ sensor (8.8 × 6.6 mm) and a 44 mm field, the required magnification is 0.2× — which is exactly the region the LTCM02-110 covers.
4. Check the sampling budget. Object-space pixel size = field of view ÷ pixel count. On the LTCM02-110 the published object resolution is 16.78 µm, so anything smaller than that cannot be measured reliably no matter how the software is tuned.
5. Confirm depth of field. Compare the published depth of field against the real height variation of the parts, not the nominal thickness. Depth of field ranges from 10 mm at 0.2× down to 0.27 mm at 2×.
6. Fix the working distance around the fixture. Working distance is a design parameter, not an adjustment. The 110 mm family is quoted at 110 mm ±3%; leave clearance for the fixture, any protective window, and the backlight assembly.

Illumination decides whether the edge exists

For dimensional measurement of a silhouette, a collimated telecentric backlight is not an accessory — it is part of the measurement system. A diffuse backlight produces a soft edge whose apparent position shifts with intensity and with the part's distance from the diffuser. A telecentric backlight produces a hard, high-contrast edge whose position is stable and independent of that distance.

Coaxial illumination is a useful dividing line when specifying. A bi-telecentric lens such as the PMS-10VP110C111T supports it natively, while an object-space cylindrical lens like the LTCM02-110 is normally paired with a separate backlight. Choose the illumination architecture first, then the lens that supports it.

How this plays out across industries

In all three, the reason a telecentric lens is specified is the same: parts are presented by a mechanism, and no mechanism repeats in Z as accurately as the measurement tolerance demands. The optics absorb that variation.

A worked station, end to end

Suppose a connector housing needs pin-position measurement with a ±20 µm tolerance, measured in silhouette, with parts fed on a rail.

If the requirement later tightens and a smaller field suffices, the same architecture scales up: the 1.0× bi-telecentric lens resolves 4.44 µm over a 14.1 × 10.3 mm field on a 1.1″ sensor, and the 2× coaxial lens reaches a 4.4 × 3.3 mm field on a 2/3″ sensor with 0.03% distortion.

Calibration and verification

Even with a 0.02% lens, calibrate the system and verify it on a schedule. Distortion below 0.1% still contributes a few microns across the field, and the scale factor depends on the exact sensor-to-lens spacing. A certified glass or chrome-on-glass target, measured at the working distance the station actually uses, establishes the scale; periodic verification catches drift from a bumped camera, a swapped lens or a changed standoff. For a full treatment of the calibration step, see the telecentric lens selection guide.

Common failure modes

• Sensor larger than the lens format. The corners fall outside the corrected field, so edge measurements at the frame boundary become unreliable while the centre still looks perfect.
• Field of view set exactly to the part size. Any placement deviation pushes the edge feature out of frame; there is no margin for processing.
• Theoretical pixel size mistaken for accuracy. Object-space pixel size is a sampling limit, not a measurement guarantee; edge detection, illumination and calibration all add error on top.
• Working distance treated as adjustable. Moving a telecentric lens away from its design working distance invalidates the magnification it was specified for.
• Supporting only the camera. Large telecentric lenses are heavy; mounting the camera and letting the lens hang from the C-mount introduces flex that shows up as calibration drift.

Frequently asked questions

1. What accuracy can a telecentric measurement station achieve?

It depends on the whole chain, but a well-specified station on a small field is normally in the single-digit micron range, with the optics contributing only a few microns when distortion is below 0.1%. Tighten the field, the sensor and the illumination together — improving only one rarely helps.

2. Do I need bi-telecentric or is object-space enough?

Object-space telecentricity is sufficient for silhouette and dimensional work with a separate backlight. Choose bi-telecentric when the system also needs a coaxial illumination path, or when working with large sensors where uniform, perpendicular imaging across the frame matters.

3. Can I measure a transparent part?

Yes, but transmission interfaces have to be resolved deliberately. Coaxial illumination through the lens, sometimes combined with a backlight, separates the surface reflection from the transmitted silhouette. Depth of field becomes the limiting factor, because the measurement plane has to sit on one interface at a time.

4. How do I know if my sensor is large enough?

Work from object-space pixel size: field of view ÷ pixel count. If that value exceeds the smallest feature you need to resolve, increase magnification or pixel count — but confirm the lens actually resolves that finely, or the extra pixels sample a blurred image.

5. Which POMEAS lenses cover my magnification range?

The telecentric lens family spans 0.2× to 2× across object-space and bi-telecentric designs, including the LTC15-110 1.5× and LTC2-150 2× models. Match the lens to your sensor format before comparing magnification.

Key takeaway: a telecentric lens makes a camera into a measuring instrument by removing position-dependent error from the image. Everything else — sensor choice, illumination, calibration and fixture — determines how much of that optical accuracy survives to the final measurement. Specify them together, and confirm every figure against the current selection manual before the design is frozen.

Further reading: the test methods behind this article are documented in POMEAS Technical Reference telecentricity test: measuring the chief-ray angle over the field (§§8.9 远心度).