Why Telecentric Measurement Drifts When the Workpiece Moves

A telecentric lens is specified for one reason: to stop the part's height from changing its apparent size. When the reading still moves as the part moves along Z, the natural conclusion is that the lens is not doing its job. Sometimes that is true. More often the drift comes from a term telecentricity was never specified to remove — defocus outside the published depth of field, a part that tilts rather than translates, an edge whose position is set by illumination, or a datum that moves with the part.

The useful first step is not to swap the lens. It is to put a number on each candidate term: telecentricity error can be bounded in micrometres from a single published angle, defocus can be checked against a published depth of field, and tilt can be bounded from geometry. Those three numbers usually separate the cases before any hardware is touched.

Why "it drifts" is not one fault

Three mechanisms produce a Z-dependent measurement, and they have different signatures.

Defocus. The part leaves the depth of field. Contrast drops, the gradient gets shallower, and a threshold-based extractor reports a different position although magnification has not changed.

Residual telecentricity. No lens is perfectly telecentric. The published figure is the residual chief-ray angle, and an axial displacement converts it into a lateral shift. The signature is a reading changing smoothly with Z while the edge stays sharp.

Tilt. The part rotates rather than translates. Telecentricity does not address foreshortening, and tilt also consumes axial range across the feature.

Separating these three before ordering anything is the point of the checklist below.

The six checks that separate lens-caused drift from everything else

1. Confirm the motion is pure translation, not tilt

Telecentricity removes the magnification change caused by axial translation. It does not remove foreshortening. A feature of length L tilted by an angle φ reads short by:

error ≈ L × (1 − cos φ)

On a 10 mm feature at φ = 1°, that is about 1.5 µm. At φ = 3° it is about 13.7 µm — already larger than many dimensional tolerances, and nothing about the lens is at fault.

Tilt also costs axial range: a feature of extent L tilted by φ spans L × sin φ in Z. A 10 mm feature at 1° spans about 0.17 mm; at 3°, about 0.52 mm. Compare that against the published depth of field.

The check is mechanical: confirm both ends of the measured feature move by the same amount in the same direction.

2. Check the axial range against the published depth of field

Telecentricity holds magnification constant *within* the depth of field. It does not extend it. If the part's Z range exceeds the published figure, the station is measuring defocused edges and the drift belongs to the first mechanism above, not to telecentricity.

3. Bound the residual telecentricity term from the published angle

For an object-side telecentric lens with a residual chief-ray angle θ, an axial displacement Δz shifts the apparent lateral position of a feature by at most:

Δy ≤ Δz × tan θ

Using the published object-side telecentricity of PMS-LTCM0061-398, <0.02°, where tan(0.02°) ≈ 3.49 × 10⁻⁴:

- Δz = 1 mm → apparent shift ≤ 0.35 µm - Δz = 5 mm → apparent shift ≤ 1.75 µm - Δz = 10 mm → apparent shift ≤ 3.49 µm

For PMS-LTCM0088-120, published at <0.04°, the same displacements give roughly 0.70 µm, 3.49 µm and 6.98 µm respectively.

If the drift you observe is much larger than these numbers, residual telecentricity is not the explanation. If it is the same order, confirm the angle for your exact serial number.

Telecentricity is not published for every model. Where the page carries the field but no value, the entry is "not published", and the bound cannot be computed at all.

4. Check whether the edge is defined by optics or by illumination

A telecentric lens does not create an edge on a low-contrast feature; it reproduces whatever contrast the illumination produces. If the backlight intensity drifts, the surface reflectivity changes, or a diffuse source is used where a collimated one is needed, the threshold crossing moves and the dimension moves with it.

The check is to hold the part still and vary only the illumination. Any movement is invisible to every lens specification.

A telecentric parallel LED backlight is the usual remedy, because it reduces the light that wraps around the edge. The VP-GLTD Series publishes beam diameters of 30 mm to 265 mm and working distances covering 45–500 mm.

5. Check the datum, not just the measured feature

A measurement is a difference between two located features. If the datum itself sits at a different height, or is a cast or moulded surface that varies, the dimension moves with the part even though both features are located correctly.

The check is to record both endpoints independently rather than the computed dimension. If one endpoint is stable and the other moves, the fault is at the moving end.

6. Confirm the mechanical stack has not moved

On a C-mount station the lens, any spacer or extension tube, and the camera flange form the optical stack. A spacer added to reach focus changes the working distance, and a loose clamp lets the assembly shift. Since a telecentric lens has a fixed working distance, a shift of the lens body is indistinguishable from a shift of the part.

The check is to measure the actual lens-to-part distance with the station running and compare it with the published working distance.

Where POMEAS telecentric lenses land on those six

Published values for models that carry an object-side telecentricity figure and a depth of field figure:

Reading the table against the six checks:

- PMS-LTCM0061-398 is the extreme case for axial range: 171.97 mm of depth of field at 398 mm, with object-side telecentricity <0.02°. - PMS-LTCM0088-120 publishes depth of field as a range, 16 mm–45 mm, rather than a single value — a reminder that the figure depends on the criterion used. - PMS-LTCM12-110 shows the other end of the trade: at 1.2× and 0.67 mm, drift with Z is far more likely to be defocus than telecentricity, despite the <0.02° figure. - LTCM02-180-G publishes <0.02° (0.1°); the page does not state what the second figure denotes. - Telecentricity is not a universal field: PMS-LTCM02-110, PMS-LTCM03-110 and PMS-LTCM05-110 carry the design but publish no value, and image-side telecentricity is published separately — PMS-LTCM0063-416-G at <0.02°.

These figures are the published values for each individual model, not measurements taken under one common condition.

What those numbers mean on the line

Scenario 1 — A 0.5 mm Z shift producing 8 µm of drift at 110 mm working distance

The station runs PMS-LTCM12-110, whose published depth of field is 0.67 mm. Parts sit within 0.5 mm of nominal, so it is not exceeded, and the bound from the <0.02° figure is only about 0.17 µm. The observed 8 µm cannot come from either term; the remaining candidates are tilt and illumination, and measuring both ends independently settles it.

Scenario 2 — A conveyor carrying parts with 12 mm of height variation

Where the height range dominates, depth of field decides the model before telecentricity does. PMS-LTCM02-180S-G publishes 12 mm at 0.2× and 180 mm with 20.13 µm resolution on a 1-inch sensor; PMS-LTCM016-189 publishes 18.75 mm at 189 mm; PMS-LTFM017-416 publishes 22.14 mm at 416 mm.

Scenario 3 — A robot-presented part with tens of millimetres of placement uncertainty

Here the axial range, not the tolerance, drives the choice. PMS-LTCM0061-398 publishes 171.97 mm of depth of field at 398 mm with 87.99 µm resolution and TV distortion 0.01% on a 2/3-inch sensor. The coarse resolution then becomes the limit: this model buys axial range, not fine feature resolution.

Scenario 4 — Drift that disappears when the light is stabilised

Drift correlated with warm-up time rather than part position, reproduced by sweeping the light controller with the part stationary, is check 4 — worth running before any optical investigation, because the fix is a stable source rather than a different lens. A VP-GLTD telecentric parallel backlight, beam diameters 30 mm to 265 mm, is the standard countermeasure for silhouette gauging.

What this article does not tell you

This is a diagnostic checklist, not a diagnosis of your station. No claim is made about what is wrong with a specific installation, and none should be inferred. Whether the drift comes from defocus, tilt, illumination or datum definition depends on measurements taken on that line.

Optical performance is not measurement accuracy. Resolution and telecentricity are properties of the lens; measurement accuracy additionally depends on pixel sampling, contrast, edge definition, illumination stability, and the algorithm.

Every figure here is the published value for the specific POMEAS model named beside it, not measured under a single common condition. Where a value is absent — depth of field for LTCM02-180-G, telecentricity for PMS-LTCM02-110 — read it as "not published".

The telecentric content already on this site covers magnification selection, working-distance families and sensor matching separately. This article separates causes after a station is built; selection questions belong to those pages.

FAQ

Does a telecentric lens completely eliminate measurement drift with part height?

No. It removes the magnification change caused by axial translation within the depth of field. It does not remove defocus outside that range, foreshortening from tilt, or edge movement caused by illumination instability.

How much error does residual telecentricity actually cause?

It can be bounded as Δz × tan θ using the published angle. For PMS-LTCM0061-398 at <0.02°, a 10 mm axial shift gives at most about 3.49 µm of apparent lateral shift.

Why is telecentricity not listed for every telecentric lens?

Because it is a measured figure that not every product page publishes. Where the value is absent, the honest reading is "not published", and the drift bound cannot be computed from the datasheet.

Can depth of field be traded against resolution?

In practice the two move together across the catalogue: PMS-LTCM12-110 publishes 0.67 mm at 1.2×, while PMS-LTCM0061-398 publishes 171.97 mm at 0.061×. Treat both as per-model values, not as a curve to interpolate.

Is tilt or defocus the more common cause of drift?

On high-magnification stations with sub-millimetre depth of field, defocus and tilt dominate because the axial budget is so small. On low-magnification long-working-distance stations, illumination and datum effects are usually larger than the telecentricity term.

Should I buy a lens with a smaller telecentricity angle to fix the drift?

Only if the bound from your current model's published angle is the same order as the drift you observe. If the observed drift is much larger, the angle is not the cause.

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.