What a Telecentric Lens Does and When You Actually Need One

A telecentric lens is bought to remove exactly one error: the change in magnification caused by a part sitting at a different height from the one the lens was set for. Better resolution, less distortion and sharper edges are not what the design guarantees. A telecentric lens can be worse than a conventional lens on resolution and cost considerably more, and still be correct, because the error it removes is often the largest term in the budget.

That makes the buying decision unusually clean, provided you compute one number first. On a conventional lens, two features at different heights are reported at different sizes, and the size of that error is a function of the height difference, the feature size and the working distance — all known before you open a catalogue. If the computed error is smaller than your tolerance, a telecentric lens spends money on a term that was never binding.

If it is larger, no resolution, distortion correction or calibration removes it, because it changes part to part with the part position.

Why the error cannot be calibrated away

The perspective term is not a fixed offset. A distortion map corrects a systematic error that is the same on every part; the perspective error depends on where the individual part sits. You can calibrate it out for one part at one height, and the next part reintroduces it.

This is why the decision is made on the geometry of the station rather than on image quality. A slightly soft image can be improved with lighting and algorithm work. A perspective term larger than the tolerance cannot.

The six questions that decide whether you need telecentric optics

1. Do the measured features sit at different heights?

This is the whole question. If every feature being measured lies in one plane, there is no height difference and no perspective term. If features sit on top of components, on different steps of a machined part, on a warped panel, or on a fixture with stack-up variation, there is.

Quantify it first: what is the largest height difference Δz between two features in the same measurement?

2. Is the perspective term larger than the tolerance?

On a conventional (entocentric) lens:

ΔM / M ≈ Δz / WD

and for a feature of size L:

error ≈ L × Δz / WD

Three worked cases using published POMEAS working distances:

Against a ±50 µm tolerance the first case fails on the perspective term alone. The third shows the same problem at the micro scale: short working distances make the term larger, not smaller, which is why high-magnification stations are so often telecentric.

Rearranged, the expression gives the working distance needed to hold the error under a budget:

WD ≥ L × Δz / allowed error

For L = 10 mm, Δz = 1 mm and a 10 µm budget, that is WD ≥ 1,000 mm — which is why pushing the working distance is not a practical fix.

3. Which kind of telecentricity you need

Three variants are specified differently, and the difference is not a quality ranking.

- Object-side telecentric — chief rays run parallel to the axis on the object side, so magnification becomes independent of the part's axial position. This is the configuration for dimensional measurement of parts with height variation. PMS-LTCM016-189 is published with object-side telecentricity <0.02° at 0.16× and 189 mm. - Image-side telecentric — chief rays run parallel on the image side, so the ray angle at the sensor is constant across the field. This matters for sensors with microlens arrays. PMS-LTCM0063-416-G is published with image-side telecentricity <0.02° at 0.063× and 416 mm. - Bi-telecentric — parallel on both sides. The most stable against both part height and sensor position, at higher cost and more bulk. PMS-LTFM017-416 is a bi-telecentric F-mount lens published at 0.17× and 416 mm with 22.14 mm depth of field and object-side telecentricity <0.02°.

4. Field of view and magnification come first, telecentricity second

Telecentricity does not set the field of view. Magnification does:

FOV = sensor width / magnification

Select the magnification from the largest region measured in one image, then check that a telecentric model exists at that magnification with the working distance and depth of field needed. PMS-LTCM016-189 publishes a 1-inch maximum sensor at 0.16×; PMS-LTCM0088-398-G publishes 1-inch at 0.088× and 398 mm.

5. Size and weight are part of the specification

A telecentric lens has a front element at least as large as the object field, so it is bigger than a conventional lens at the same magnification, and bi-telecentric designs are larger again. Published lengths: PMS-LTCM016-189 at 289.09 mm, PMS-LTFM1-127 at 297.35 mm, and PMS-LTFM017-416 at 712 mm with a Φ254 mm object field.

On a retrofit this is often the constraint that decides, before any optical consideration.

6. When a conventional lens is the right answer

A telecentric lens is the wrong choice in several common situations.

- All measured features lie in one plane. No height difference, no perspective term. - The tolerance is comfortably larger than the perspective term. Compute it before spending anything. - The task is presence, absence, counting, or code reading rather than dimensioning. - The station needs variable magnification. Telecentric lenses are fixed-magnification by design. - The part is measured with a contact or non-imaging sensor anyway, and the camera only locates it.

A conventional lens with a longer working distance is also a legitimate answer when the perspective term is marginal rather than dominant — the formula tells you whether that is true.

Where POMEAS telecentric lenses land on those six

Reading the table against the six questions:

- PMS-LTCM016-189 is the clearest object-side case: <0.02° published, 18.75 mm depth of field, 25.17 µm resolution, 1-inch maximum sensor, 289.09 mm long. - PMS-LTCM0063-416-G publishes image-side telecentricity <0.02°, a maximum 4-inch sensor, a 120.73 mm field and 63.8 µm resolution at 0.063× and 416 mm. - PMS-LTFM017-416 and PMS-LTFM1-127 are bi-telecentric F-mount designs: 22.14 mm depth of field and a 241.2 mm row field for 8K × 5 µm line-scan, and 10.82 µm resolution with 1.29 mm depth of field at 1.0× and 127 mm. Their 712 mm and 297.35 mm lengths are the first thing to check against the envelope. - PMS-LTDC00880176-120 carries two channels, 0.088× and 0.176×, sharing a 120 mm working distance through a dual C-mount, with depth of field 16 mm / 7.2 mm and F 4.46 / F8. - PMS-LTCM0088-398-G publishes 1-inch coverage at 398 mm with distortion 0.01%, NA 0.0074 and F 6, but no resolution, depth of field or telecentricity.

Each row carries only what that model publishes. Where a value is missing the reading is "not published", to be requested rather than assumed.

What those numbers mean on the line

Scenario 1 — Machined housing with features on two levels

Features on the top face and on a recessed shoulder 1.5 mm below are measured in one image at 189 mm working distance. The perspective term for a 10 mm feature is about 79 µm, against a ±30 µm tolerance. PMS-LTCM016-189 at <0.02° removes the term, and its 18.75 mm depth of field keeps both levels in focus.

Scenario 2 — Flat stamped part, all features coplanar

Every measured feature sits on one surface, the part is held flat, and the tolerance is ±100 µm. The perspective term is effectively zero: a telecentric lens would remove an error that is not present.

Scenario 3 — Large panel on a line-scan camera

The part is larger than any area sensor field and the station runs a line-scan camera. PMS-LTFM017-416 publishes a 241.2 mm row field for 8K × 5 µm with a Φ254 mm object field, 22.14 mm depth of field and bi-telecentric optics at 416 mm. The bi-telecentric design keeps magnification constant across the full width of the web.

Scenario 4 — One station, two jobs

A cell has to locate a large part and then measure a small feature on it, and there is no room for two lenses. PMS-LTDC00880176-120 provides 0.088× and 0.176× from one body at a shared 120 mm working distance with dual C-mount.

What this article does not tell you

This is a technical and selection reference, not a diagnosis of your station. No claim is made that your application requires telecentric optics, and none should be inferred. That depends on the height range present, the tolerance, the illumination and the algorithm.

Optical performance is not measurement accuracy. Telecentricity removes one term; measurement accuracy additionally depends on pixel sampling, contrast, edge definition, illumination stability, and the algorithm. A telecentric lens removes one reason why a station was inaccurate.

Every figure here is the published value for the specific POMEAS model named beside it, not measured under a single common condition. Telecentricity in particular is published for some models and not others: PMS-LTCM0088-398-G and PMS-LTFM1-127 carry no figure, and none should be assumed.

The telecentric content already on this site covers the four-number selection procedure, per-magnification model families, working-distance groups and sensor matching as separate pages. This article is the overview; those carry the selection numbers.

FAQ

What does a telecentric lens actually fix?

One thing: the change in magnification that occurs when a part sits at a different height. It does not improve resolution and does not extend depth of field.

How large is the error it removes?

Approximately L × Δz / WD. At 110 mm working distance, a 1 mm height difference on a 10 mm feature gives about 91 µm.

What is the difference between object-side and bi-telecentric?

Object-side telecentricity stabilises magnification against part height; bi-telecentricity additionally stabilises the ray angles arriving at the sensor. Bi-telecentric designs cost more and are larger.

Does telecentricity increase depth of field?

No. Depth of field is a separate specification. A telecentric lens keeps magnification constant within its depth of field; it does not extend that range.

When should I not use a telecentric lens?

When all measured features lie in one plane, when the tolerance is larger than the perspective term, when the task is presence or code reading rather than dimensioning, or when the station needs variable magnification.

Can I calibrate the perspective error out instead?

Only for a fixed part at a fixed height. The error varies with where each individual part sits, so it is not a systematic offset a distortion map can remove.

Next Step

If you are specifying a replacement and need measured values for your own conditions — your sensor size, your working distance, your feature size — send us those four numbers and we will run them against the current catalogue and tell you what actually fits, including when nothing does.