Start With the Measurement Error You Need to Control

A telecentric lens is not a general cure for measurement variation. The selection process should start with an object-space error budget expressed in the same units as the drawing tolerance: millimetres or micrometres at the part, not pixels alone.

Does part height move the measured edge?

With common entocentric optics, displacement toward or away from the lens changes apparent magnification and perspective. An object-space telecentric design makes image size substantially more stable while the object remains within its specified telecentric range. The Opto Engineering telecentric-lens tutorial identifies object displacement, perspective, distortion and edge uncertainty as relevant limits in dimensional measurement.

Test the real boundary condition. Include part thickness tolerance, conveyor runout, fixture variation and any deliberate Z-axis movement. If their combined effect moves an edge beyond its allocated optical-error allowance, object-space telecentricity has a defensible purpose.

Separate perspective, sensor and illumination errors

A changing silhouette with object height suggests object-side geometry. A shift after camera or sensor mounting changes points toward image-side alignment or calibration. Edge movement that follows illumination angle, intensity or threshold settings is not automatically an optical telecentricity problem.

This separation matters commercially. A larger telecentric lens can add mass, envelope and cost because its front optical components must accommodate the object field. It cannot correct an unstable fixture or poorly defined edge.

Allocate tolerance before choosing hardware

Break total measurement uncertainty into practical contributors: calibration residual, optical distortion, magnification variation over depth, edge localisation, sampling, fixture repeatability and environmental drift. The allocation need not be equal. It does need to leave margin for integration and production variation.

Engineering inference: if a feature tolerance is ±0.020 mm, assigning the entire ±0.020 mm to the lens is not a workable specification. The optical allowance must be only one defined portion of the system budget; its value depends on the other contributors and the required acceptance risk.

Telecentric Lens Selection Decision Tree

1. Does object-height variation produce unacceptable apparent size or edge movement? Evaluate an object-space telecentric lens.
2. Is controlled sensor-side chief-ray behaviour or tolerance to image-plane positioning a material requirement? Evaluate an image-space telecentric lens and obtain supplier data for the intended sensor.
3. Do object-depth variation and image-side stability both consume meaningful error allowance? Compare a bi-telecentric lens with separate mitigation of the two errors.
4. Does telecentricity leave the dominant error unchanged? Improve calibration, illumination, fixturing or sampling first, and compare a conventional lens.

Telecentric Lens Type Selection for Dimensional Metrology

The table is an engineering selection synthesis. Architecture names do not establish performance; verify model-specific limits with the optical supplier.

Comparing the Three Telecentric Architectures

Object-space telecentric lens

This is the usual starting point when a part’s Z position changes but its measured X-Y dimensions must remain stable. It is well suited to silhouette measurement, outside diameters and edges on components that cannot be held at one exact height. Performance still depends on telecentricity, distortion, focus, illumination and the qualified depth range.

Telecentricity should not be interpreted as an unlimited zone of perfectly constant magnification. The verified manufacturer guidance describes telecentric slope as the more meaningful measure of displacement-related error. Buyers should therefore request error versus object displacement, not only a nominal “telecentric depth” label.

Image-space telecentric lens

Image-space telecentricity controls ray geometry at the sensor side. It can be relevant where chief-ray angle, image-plane position or sensor integration imposes a specific constraint. It does not, by itself, establish that varying object height will preserve measured dimensions.

Specification caution: the supplied verified sources do not provide universal image-space tolerance values. Treat any expected benefit as design-specific and request chief-ray, image-plane and sensor-format data for the complete camera-lens combination.

Bi-telecentric lens

A bi-telecentric system combines object-side and image-side telecentric behaviour. That can be justified when object depth and sensor-side stability both have allocated error. It also introduces packaging and purchasing trade-offs, so “bi-telecentric” should not be used as a substitute for numerical acceptance criteria.

Buyers can review available telecentric lenses for metrology after identifying which side of the optical system requires control. Compare quoted performance at the required field, wavelength, aperture and working distance rather than by architecture name alone.

Size the Optical Layout for Field of View and Sensor Coverage

Required magnification is approximately the active sensor dimension divided by the corresponding object field dimension, provided both dimensions use the same units. Calculate horizontal and vertical values; the more restrictive axis governs. The broader workflow for field of view, sensor, and distance selection helps expose mismatches before mechanical design is frozen.

A short sampling example

Suppose a 12 mm-wide active sensor must cover a 24 mm-wide object field. The planning magnification is 12/24 = 0.5×. With 3.45 µm pixels, one pixel corresponds nominally to 3.45/0.5 = 6.9 µm at the object.

That figure is sampling pitch, not guaranteed measurement accuracy. A distributor’s telecentric lens selection guidance recommends planning around four pixels across a minimum feature rather than treating two-pixel sampling as sufficient in practice. Under that heuristic, the example provides about 27.6 µm across four pixels. Edge algorithms may estimate subpixel positions, but the validated result still depends on contrast, noise, calibration and edge shape.

Coverage is more than nominal sensor format

Confirm the image circle covers the active sensor and that performance remains acceptable at field corners. Review distortion, telecentricity and contrast across the usable field, not only on axis. Telecentric Lens Selection can fail when nominal field coverage is accepted without checking whether the outer field remains inside the measurement budget.

For line scan systems, width is determined by pixel count and pitch rather than a two-dimensional area array. Use line scan sensor pixel count, pitch, and lens coverage to establish the required image width, then obtain telecentricity and distortion data across the full scan line.

Working Distance, Depth Range and Mechanical Clearance

Telecentric lens working distance is a model-specific design condition, not an installation range that can be shifted freely. Verify where the supplier measures it, the associated magnification, and whether protective windows or covers are included in the optical path.

Mechanical clearance can eliminate an otherwise suitable lens. Reserve space for fixtures, conveyors, robot access, reject mechanisms and illumination. Because telecentric front elements must accommodate the object field, wide-field models can be physically large. Check diameter and mass early, including mount stiffness and centre-of-gravity effects.

Depth of field and telecentric range are also different acceptance questions. A feature may remain visibly sharp while magnification change exceeds the metrology allowance, or remain geometrically stable while contrast is inadequate for repeatable edge extraction. Request performance over the complete object-height range and test the production illumination configuration.

Specifications That Determine Repeatability

Telecentricity, distortion and magnification stability

These values must be assessed together. Low displacement-related magnification error does not remove geometric distortion across the field. The verified manufacturer tutorial notes that distortion is a significant measurement limitation and that even a one-pixel difference can matter in some applications. Calibration may reduce systematic distortion, but the residual and its stability must be validated.

Ask how each specification is defined, under what wavelength and aperture it was measured, and whether the value is maximum, typical or local. Two datasheets can use similar labels while reporting different conditions.

Resolution, contrast and aperture

A lens must transfer enough contrast at the spatial frequencies relevant to the sensor and edge. Pixel count alone cannot establish this. Aperture changes light throughput and depth of field and may alter diffraction or aberration balance. The guide to lens specifications including F-number and MTF provides the terms needed to compare supplier data.

Match wavelength and coatings to the actual source. Backlighting commonly supports clean silhouettes, while reflective or transparent parts may require a different geometry. Record wavelength, diffuser, polarisation and strobe conditions because edge position can change when illumination changes.

Mechanical review should cover sensor format, mount, flange constraints, focus locking, vibration, cable clearance and environmental exposure. An optically suitable assembly that cannot hold alignment is not a metrology solution.

When Telecentric Performance Will Not Solve the Problem

Object-space telecentricity may be unnecessary when part-height variation is so small that conventional-lens magnification change remains comfortably below its allocated error. This must be demonstrated over production tolerances rather than assumed from a nominal fixture height.

Other failure modes deserve priority when they dominate:

• too few pixels across the smallest relevant feature;
• low or inconsistent edge contrast;
• fixture motion in X, Y or rotation;
• calibration targets or procedures that do not match the working volume;
• sensor noise, vibration or thermal drift;
• distortion residuals larger than the depth-related error being corrected.

A standard machine vision lens can therefore be the more practical choice for fixed-height parts, moderate accuracy requirements or systems where calibration removes stable geometric error. The buyer-facing consequence is lower optical complexity and often a smaller package, provided validation shows that the measurement budget is still met.

Build the Specification Package Before Requesting a Quote

A defensible Telecentric Lens Selection request gives the supplier enough information to calculate feasibility and identify missing constraints.

Part and measurement inputs

• maximum object width and height, including positional variation;
• minimum feature or edge spacing;
• part-height range and fixture runout;
• drawing tolerance and allocated optical error;
• reflective, transparent or silhouette edge condition.

Camera and optical inputs

• sensor model, active dimensions, pixel count and pixel pitch;
• required field of view and orientation;
• frame rate or line rate;
• wavelength, illumination geometry and available exposure;
• required mount and any cover glass.

Mechanical and environmental inputs

• available working distance and permitted lens envelope;
• fixture, conveyor and robot clearances;
• vibration, temperature and contamination conditions;
• focus-locking and alignment provisions.

Request a custom optical assembly when no standard model simultaneously meets field, depth, sensor, wavelength and packaging constraints. Customisation is most productive after acceptance criteria are numerical; otherwise the project risks optimising a lens without defining system-level success.

For selection support, submit the part dimensions, height variation, field of view, sensor format, working distance, tolerance allocation, illumination wavelength and mechanical envelope. Those inputs allow a supplier to compare conventional, object-space, image-space and bi-telecentric options against the same measurement problem.