Telecentric Lenses for Automotive Parts: Dimensional Inspection Selection

Automotive dimension inspection has a specific shape: the same feature family is measured on every part, all shift, every shift; the tolerance is fixed by the drawing, and the measurement result has to survive gauge studies and customer audits. The parts themselves are rarely kind to optics. A stamped bracket mixes a thin web with a tall flange. A machined housing has bores at different depths. A molded connector shell presents edges at several heights in one field of view. On all of these, the question is not whether the camera sees the edge, but whether the edge position it reports is the same when the part sits 2 mm higher in the fixture. That is exactly the property a telecentric lens is built to deliver, and it is the reason telecentric optics are the default choice for gauge-grade dimensional stations in the automotive supply chain.

This page is written for the practical selection case: a part family, a tolerance, a cycle time and a fixture. It covers why telecentric projection changes the error budget on stepped automotive parts, how to set magnification and field of view from the sensor and the drawing, a shortlist of published model figures for common selections, lighting choices for metal and coated surfaces, and the boundary where a telecentric camera station should hand over to an inline measurement head or a flash measuring system instead.

Why Telecentric Projection Matters on Automotive Parts

A conventional machine-vision lens views the object through a cone of rays: each point on the part is imaged along a different angle, and the angle changes with position in the field. If the part moves toward or away from the lens - a fixture that wears, a bracket that flexes, a casting datum that varies by a fraction of a millimetre - the magnification effectively changes and the edge appears to move even though the part dimension has not. On a flat part measured flat, the effect is small. On a stepped automotive part with a 5-10 mm height difference between two measured edges, the apparent shift can be a significant fraction of the tolerance, and it does not average out: it is a systematic error that changes with every fixture variation.

A telecentric lens collects only the rays inside a narrow cone parallel to the optical axis on the object side. Move the part within the depth of field and the magnification stays constant, so the edge stays where it is. The strength of that property is published per model as object-side telecentricity, given in degrees. For the PMS-10110CVP111T 1.0X bi-telecentric lens, object telecentricity is <0.05-0.1 degrees (ProdParameters, ProdID 251). The apparent edge shift over a height step is roughly the step multiplied by the tangent of that angle (derived): over a 10 mm step, 0.05-0.1 degrees corresponds to about 9-17 micrometres of edge movement - a figure that either fits inside the drawing tolerance or does not, and can be checked before any hardware is bought. For the PMS-LTCM016-189 and PMS-LTCM0061-398 high-resolution models, object telecentricity is <0.02 degrees (ProdID 426 / 427), which reduces the same 10 mm step to roughly 3.5 micrometres (derived).

The second published property that matters on automotive parts is TV distortion. Distortion bends straight edges into curves, and on a long stamped edge or a large bore wall the middle of the field no longer matches the ends. The standard and coaxial telecentric series publish TV distortion of 0.01-0.02% (ProdID 329 / 398 / 426 / 427) and the bi-telecentric 1.1-inch model <0.05-0.1% (ProdID 251). On a 20 mm measured span, 0.02% corresponds to 4 micrometres of bow (derived) - small enough to be calibrated out once, and stable afterwards because the distortion of a fixed-focal telecentric design does not change with focus or magnification.

What a Standard FA Lens Costs You on Stepped Parts

The comparison case is common in automotive plants: a FA lens is already on the shelf, the part tolerance looks loose, and the temptation is to reuse it. Three things break. First, perspective error on height steps, as above: the reported edge position depends on where the part sits in Z, and the dependence is not calibratable away because it varies per part. Second, magnification drift with focus: refocusing after a fixture change silently rescales the pixel-to-millimetre conversion, which shows up in gauge R&R as an operator or shift effect rather than an optical one. Third, distortion that is neither published nor specified for the generic FA lens, so the error budget cannot even be written down honestly.

If the part is flat, single-height, and measured with a generous tolerance, a calibrated FA lens can be adequate - the honest boundary question is answered in our guide to what a telecentric lens does and when you actually need one (what-a-telecentric-lens-does-and-when-you-need-one.html). If the part has measured features at more than one height, or the tolerance is below roughly 0.05 mm, the projection property stops being optional. A related failure mode appears when an edge is measured against a calibrated image but the calibration was made at a different working distance; the working distance of a telecentric lens is a set optical property, not a mounting preference (telecentric-lens-working-distance-and-how-it-is-set.html).

Setting Magnification and FOV from the Part Drawing

The selection arithmetic is short and should be done before looking at model lists. Start from the drawing: the largest feature family that must appear in one image sets the object-side field of view, with margin for part placement tolerance in the fixture. The field of view follows from the sensor and the magnification: FOV = sensor dimensions divided by magnification (derived). A 2/3-inch sensor (about 8.8 x 6.6 mm, industry-standard format, derived) at 1.0X covers about 8.8 x 6.6 mm of object; the same sensor at 0.2X covers about 44 x 33 mm (derived), which is why the low-magnification telecentric models carry the larger part families.

Then check sampling. The object-side pixel size is the sensor pixel pitch divided by the magnification (derived). With a common 3.45 micrometre pixel sensor at 1.0X, the object sampling is 3.45 micrometres per pixel; at 0.2X it is 17.25 micrometres (derived). For a gauge application, plan on the published lens resolution being finer than the object pixel size, and on at least 10 pixels spanning the tolerance band as a practical rule of thumb (heuristic, not a published figure) so that edge algorithms have room to work. A tolerance of plus/minus 0.05 mm means a 0.1 mm band; at 17.25 micrometres per pixel that is about 6 pixels - thin - which pushes the selection either to a higher magnification with a smaller FOV, or to a model with finer resolution such as the 1.0X bi-telecentric at 5.37 micrometres published resolution (ProdID 251, best case of its published 5.37-13.42 micrometre range).

Sensor matching is part of the same step: the image circle of the lens has to cover the sensor, and the published maximum sensor size per model is the hard boundary. Matching rules are set out in the guide to telecentric lens sensor size and image circle (telecentric-lens-sensor-size-and-image-circle-matching.html).

Published Model Shortlist for Common Automotive Selections

All values below are quoted from the current product parameter records (read back 2026-11-09); nothing in the table is estimated. Where a value is not published for a model, it is marked not stated rather than filled in.

How to read the table against typical automotive part families (grouping is engineering interpretation, the figures are published):

• Small turned and machined features - pins, shafts, bore plugs, threaded inserts. The 1.0X class with a 2/3-inch or 1.1-inch sensor covers single features of up to roughly 8.8 x 6.6 mm (derived for 2/3") or the published dia 18.4 mm object FOV of the bi-telecentric model (ProdID 251). The coaxial variant (ID 329) adds built-in coaxial illumination for measuring machined faces where the edge is a step between surfaces rather than a silhouette.
• Stamped and formed metal parts - brackets, clips, contact plates. Height differences push the choice to the low-magnification class: 0.2X at WD 180 mm with a 64 x 48 mm maximum FOV (ID 232), or 0.16X at WD 189 mm with 18.75 mm depth of field (ID 426) when the part is tall rather than wide.
• Bulkfixtured large parts measured in one image - housing end faces, cover plates. The 0.061X model at 398 mm WD (ID 427) gives 171.97 mm depth of field, which absorbs large part-to-fixture variation, at the cost of 87.99 micrometre published resolution - adequate for 0.5 mm-class tolerances (interpretation), not for gauge-grade work.
• Feature family too large for one FOV. Either move to a lower-magnification model with a larger object field, or split the part into measured zones with separate stations; stitching is possible but multiplies the calibration burden and is usually the wrong trade on a gauge station.

Full published parameters for the workhorse 1.1-inch bi-telecentric model are on its product page (1.0X 1.1 Inch Bi-Telecentric Lens 110mm WD), and the equivalent 1.0X C-mount standard model is here (LTCM1-110 product page). Where the part family varies between small and large, a telecentric macro zoom such as the PMS-LZME-0510M (0.5X-1.0X, 2:1 zoom ratio, WD 75-106 mm depending on magnification - ProdParameters, ID 249) covers a range with one station: telecentric macro zoom lens.

Lighting: Backlight, Coaxial and Coated Surfaces

On automotive metal parts, the measurement edge is usually a silhouette, and silhouette accuracy is decided by the light more than by the lens. A collimated backlight behind the part keeps the telecentric cone intact and produces a sharp, stable edge; a diffuse panel bleeds light around the edge and widens it by a predictable-but-unwelcome margin. The trade-offs are set out in the backlight selection guide (diffuse-vs-collimated-backlight-measurement.html). For edges between two machined faces - a shoulder, a groove, a counterbore - the coaxial port on models such as ID 329 projects light down the optics and reads the step directly. Small holes and apertures have their own edge failure mode under backlighting, covered in the micro-hole edge guide (why-the-edge-of-a-micro-hole-blurs-in-backlighting.html).

Lacquered, plated and glass-filled surfaces appear in exterior and lighting-component inspection, and each adds reflections that can ghost an edge. Telecentric optics reduce one class of the problem by removing perspective doubles, but surface ghosts need lighting-side treatment; the failure patterns and fixes are in the transparent-parts guide (ghost-edges-on-glass-and-transparent-part-inspection.html). Operating temperature is a real automotive-constraint check: the standard series is published at -10 to +50 degrees C working temperature (ProdID 329 / 398 / 426), while the bi-telecentric 1.1-inch model is specified at 0-60 degrees C (ProdID 251) - a plant floor near a paint line or an oven is a selection input, not a footnote.

Where a Telecentric Station Ends: Inline Heads and Flash Systems

A telecentric camera station is the right shape when one part family is measured at one station, at a controlled pace, with an operator or robot loading the fixture. Two automotive cases need a different product, and naming the boundary honestly saves a mis-specified order.

First, when the measurement has to happen on the transfer path itself - every part, in motion, no fixture - the telecentric station is replaced by an online measurement head. The HM series heads mount over the line, measure the silhouette of every passing part and report per cycle; the selection logic for the five head models, ranges and their limits is in the HM series guide (inline-dimensional-measurement-system.html and hm-series-online-image-measuring-head-selection.html), with integration and sorting logic in the online gauge guide (online-gauge-plc-integration-and-sorting.html). The product entry point is the HM telecentric measurement system page (HM in-line measurement system).

Second, when the part family needs dozens of dimensions per part - a machined housing with bores, grooves, depths and true-position callouts - the one-dimension-per-image pace of a telecentric station becomes the bottleneck. A flash measuring system captures the full field in one shot and outputs the whole dimension set at once; the boundary of what one-shot measurement covers is in the multi-dimension limits guide (one-shot-multi-dimension-measurement-limits.html), and the IMAGE 3 system guide covers the full platform (image-dimension-measurement-system-guide.html).

Selection Checklist for an Automotive Telecentric Station

• Write down the largest feature family and its tolerance band from the drawing; compute FOV and object pixel size from sensor and magnification (derived arithmetic above) before contacting a vendor.
• Check the height step between measured features against the published object telecentricity of the candidate model; convert step x tan(angle) into micrometres (derived) and compare with tolerance.
• Check TV distortion across the measured span the same way.
• Fix the lighting path with the lens: collimated backlight for silhouettes, coaxial for machined-face steps; specify both in the station drawing, not afterwards.
• Check the working temperature specification against the plant location of the station.
• Decide the boundary honestly: every part on a moving line points to an inline head; dozens of dimensions per part points to a flash system; one station, one family, gauge-grade tolerance points to the telecentric station.

Why does a standard FA lens give inconsistent dimensions on the same stepped automotive part?

Because a conventional lens images each field point along a different ray angle, the apparent position of an edge depends on the height of that edge in Z. When a bracket flange sits 2 mm higher than the calibration part, the flange edge is reported at a shifted position even though the dimension has not changed. The shift is systematic, varies part to part with fixture state, and cannot be calibrated away. A telecentric lens removes the mechanism by imaging through a cone parallel to the axis, holding magnification constant across the depth of field.

Which magnification should I pick for a 0.1 mm tolerance band on a machined part?

Work backwards from sampling: object pixel size = sensor pixel pitch divided by magnification (derived). With a 3.45 micrometre pixel sensor, 1.0X gives 3.45 micrometres per pixel and about 29 pixels across a 0.1 mm band (derived), which is comfortable for edge algorithms; 0.2X gives 17.25 micrometres and about 6 pixels, which is thin for gauge work (rule of thumb: aim for at least 10 pixels per tolerance band, heuristic). So the choice is the highest magnification whose field of view still covers the part family, not the other way round.

What field of view does a 2/3-inch sensor cover at 1.0X, and is that enough for my part?

A 2/3-inch sensor is about 8.8 x 6.6 mm (industry-standard format, derived), so at 1.0X the object field is about 8.8 x 6.6 mm (derived). Whether that is enough depends only on the largest feature family that must appear in one image plus fixture placement margin. If the family is larger, step down in magnification: the published options in the current range run from 1.0X models with 110 mm working distance to 0.061X at 398 mm WD with 171.97 mm depth of field (ProdParameters, ID 427).

Backlight or coaxial light for automotive metal parts on a telecentric station?

For silhouette edges - outer profiles of stamped and machined parts - use a collimated backlight; it preserves the parallel ray cone that makes the telecentric edge accurate, and a diffuse panel instead widens and softens the edge. For steps between two visible faces - shoulders, grooves, counterbores - use the built-in coaxial illumination of models such as PMS-LTC1-110C (ProdID 329). Many stations carry both and switch per feature group. Small holes and apertures need attention to the backlight geometry, since their edges blur under incorrect collimation.

When should an automotive line move from a telecentric station to inline heads or a flash system?

Move to an inline measurement head when the measurement must happen on the transfer path for every part in motion without a fixture - the HM heads report one silhouette result per cycle and can stop or reject on the signal. Move to a flash measuring system when one part needs dozens of dimensions per cycle - bores, grooves, depths, positions - because a telecentric station measures one image per dimension and becomes the cycle bottleneck. Stay with the telecentric station for one part family, one station, gauge-grade tolerance and manageable cycle time.