A part that sits 1 mm nearer to the lens looks larger. On a 20 mm part measured at 200 mm working distance, it looks 0.1 mm larger — a 0.5% size error that calibration cannot remove, because it moves with each part's position. This guide gives the two-line formula behind that number, works it through a nest-tolerance case, and separates what telecentric lenses remove from what they leave behind. You will finish able to read your own tolerance and fixture variation, and to decide between a standard machine vision lens and a telecentric one.
Why a standard lens changes size when the part moves
A standard lens looks at every point of the part along a chief ray aimed through the entrance pupil. Image size therefore scales with distance: apparent size is proportional to 1/s, with s the pupil-to-part distance. Move the part a millimetre closer on that stand-off and magnification rises by 200/199 — about 0.5%. That is perspective parallax, and every entocentric lens has it; the selection consequences are worked through in entocentric vs telecentric lenses: when to use each, and the base vocabulary sits in machine vision optics basics: FoV, WD, magnification and f-number.
The same geometry gives the convergence angle. A 20 mm part seen from that distance subtends atan(10/200) = ±2.9° at the pupil, so a depth change ds slides each edge point sideways by ds × tan(angle), and the measured width moves by H × ds/s. Racking a standard lens to reframe shifts magnification by the same mechanism — how camera field of view changes with working distance plots that curve.
Worked example: a 1 mm nest shift on a 20 mm part at 200 mm
Inputs, from a typical bench gauging case:
- Dimension gauged: H = 20 mm.
- Calibrated working distance: s = 200 mm.
- Nest depth variation: ds = ±1 mm.
- Optics: f = 12 mm on a 1/3" camera, sensor 4.8 mm wide.
Rule from the last section: relative error = ds/s. First confirm the magnification:
The part spans 20 × 0.064 = 1.28 mm at the sensor and the field of view is 4.8 / 0.064 ≈ 75 mm, so it fits with margin. Now let the part sit at the near end of the nest, s − ds = 199 mm:
The error is +0.10 mm, or +0.5% of the dimension (0.1005 / 20 = 0.5%). At the far end, 20 × 200 / 201 = 19.90 mm — the same ±0.10 mm in the other direction. At the sensor, ±0.10 × 0.064 = ±6.4 µm, about 2 pixels of a 2 MP camera across a 75 mm field: the error shows in the image rather than hiding in noise.
Decision. The relative rule reads it straight off: ds/s = 0.5%. If the drawing tolerance on this dimension is ±0.05 mm (±0.25% of 20 mm), the perspective error at ±1 mm of nest slack is twice the tolerance. Two ways out: fixture the depth ten times tighter, which cuts the error to 0.05%, or a fifth of the tolerance — or fit a telecentric lens and leave the nest as it is. If the tolerance were ±0.5 mm (±2.5%), the standard lens stays: calibrate at mid-nest and the ±0.5% error holds a 5:1 margin. Budget rule: keep H × ds/s under a third of the tolerance; which side of that line your numbers land on decides the lens.
What telecentricity is: chief rays parallel to the axis
Object-side telecentricity places the aperture stop at the front focal plane. Every chief ray then leaves the lens parallel to the optical axis, and every point of the part is seen head-on. Apparent size stops varying with object distance: magnification is constant across the lens's telecentric range, so the nest shift of the example changes nothing but focus. That constancy is the entire value, and it cannot be calibrated into a standard lens.
A bi-telecentric lens repeats the property on the image side, so magnification also ignores sensor spacing and tilt changes; when image-side parallelism matters is covered in object space vs image space telecentricity.
The residual: reading the telecentricity specification
Real chief rays are only nearly parallel. Datasheets give telecentricity as a chief-ray angle, commonly 0.1° to 0.5° for measurement lenses. A point sitting Δs from the calibrated plane shifts sideways by Δs × tan(angle):
At θ = 0.1°, the ±1 mm nest shift of the example moves each point 1 mm × 0.00175 = ±1.7 µm; at 0.5° it is ±8.7 µm. Against the ±100 µm of the standard lens, that is 10–60× smaller — micrometres instead of a tenth of a millimetre. Depth and field position are separate effects: distortion changes magnification across the image while telecentricity governs it along depth, and the two stay on separate datasheet lines in telecentric lens distortion and telecentricity explained.
Telecentric range is not depth of field
Two depth limits apply, set by different things. The telecentric range is the depth band over which magnification stays constant — set by the chief-ray geometry above. Depth of field is the band over which blur stays under the circle of confusion — set by the f-number. The part must sit inside both, and the tighter band governs the fixture design.
Depth of field scales roughly linearly with f-number: closing from F2.8 to F8 multiplies it by 8 / 2.8 ≈ 2.9. At fixed f-number it shrinks roughly with the square of magnification, which is why 1x telecentrics feel shallow next to 0.166x ones. Work your numbers in the depth of field calculator, see the validation routine in telecentric depth of field: validate measurement accuracy, and the iris trade-off in lens aperture, working distance and depth of field.
When the standard lens stays
Perspective error matters only against a tolerance. At the calibrated stand-off of the example, each millimetre of depth slack costs 0.5% of measured size; the table reads off from your own numbers:
| Task | Depth slack at s = 200 mm | Error on a 20 mm part | Lens decision |
|---|---|---|---|
| Code reading, OCR, presence | ±1 mm | ±0.10 mm (0.5%) | Standard lens |
| Defect classification | ±2 mm | ±0.20 mm (1%) | Standard lens |
| Gauging to ±0.5% of size | the ±1 mm shift | ±0.10 mm (0.5%) | Tighten the nest or go telecentric |
| Gauging to ±0.25% of size | the same shift | ±0.10 mm | Telecentric |
Distance buys the error down in direct proportion: the same nest slack at a 440 mm stand-off is 1/440 = 0.23%. Where the geometry needs stand-off and constant magnification together — tall guards, hot parts you cannot approach — a long-working-distance telecentric holds both: the LS-TCL-1972 images at 1x with 440 mm working distance on a 2/3" sensor, in a 177.5 mm barrel.
- Sensor format
- 2/3"
- Mount
- C-mount
- Image circle
- 8-11 mm
- Type
- Object-side telecentric
Choosing magnification, field of view and backlight
Telecentric magnification is fixed, so choose it from the sensor: field of view = sensor width ÷ magnification, with 10–20% margin beyond the part so edges and nest walls stay inside the field. Object-space pixel size follows as field ÷ pixel count, worked through in calculate object space pixel size from magnification; the sensor-side version of the same conversion is in telecentric lens field of view calculation.
The three catalog lenses below span the usual range: 0.5x at 110 mm working distance on a 1" sensor, 1x at 108 mm on a 2/3" sensor, and 0.166x at 258 mm on a 1" sensor:
| Specification | LS-TCL-2194 | LS-TCL-2118 | LS-TCL-1422 |
|---|---|---|---|
| Magnification (x) | 0.5 | 1 | 0.166 |
| Working distance (mm) | 110 | 108±2 | 258 |
| Sensor format | 1" | 2/3" | 1" |
| Max sensor size (dia. mm) | 16.0 | 11.0 | 16.6 |
| Image-side F/# | 8 | 7 | 6.5 |
| Image-side MTF30 (lp/mm) | 116 | 114 | – |
| Object-side resolution (um) | 10.7 | 4.7 | – |
| Numerical aperture NA | 0.03123 | 0.0712 | – |
| Image-side distortion (% max) | 0.03 | 0.04 | – |
| Object-side telecentricity (deg max) | 0.08 | 0.05 | – |
| Total length TTL (mm) | 274 | 105.1 | – |
| Front filter thread | M44x0.5 | M30x0.5 | – |
| Price | $1,539.00 | $1,239.00 | $3,099.00 |
All 3 share: Mount C-mount. Values as printed in each product's data sheet; open a part for its drawing and full specifications.
Reading the spread: at 0.166x, a 12.8 mm-wide 1" sensor covers 12.8 / 0.166 ≈ 77 mm, so the 20 mm part of the worked example keeps a 3.8× width margin; at 1x the same sensor covers 12.8 mm and the part fills it. More magnification buys pixels on the part and shrinks depth of field with its square. Fields beyond a 1" format move up in sensor size: the LS-TCL-0885 is bi-telecentric at 0.287x with 288 mm working distance for APS-H sensors over M58, rated beyond 25 MP.
- Sensor format
- APS-H
- Mount
- M58
- Resolution
- Over 25 MP
- Image circle
- 30-45 mm
Two pairings complete the choice. The front barrel must see the whole field, so its diameter grows with field size rather than sensor size — the sizing consequences are in telecentric lens size: front optic diameter sets the field of view. And for edge-based dimensions, backlight with collimated or telecentric light: a diffuse panel floods the edge with rays at many angles and fattens the silhouette, as compared in collimated backlight or diffuse backlight for edge gauging.
Checks before you order
- Compute the field as sensor width ÷ magnification and keep part plus 10–20% margin inside it. Telecentric fields do not reframe — racking the lens only moves focus.
- Lay the nest's real depth band over both limits, telecentric range and depth of field, and design the nest to the tighter one.
- Check the front barrel against the part's largest cross-section, not against the sensor format.
- Match mount and image circle to the sensor: C-mount through M58 across the catalog, image circle at or above the sensor diagonal.
- Put distortion and the depth-error term on one accuracy budget next to the tolerance; the larger of the two sets the design.
Shortlist against your geometry in the telecentric lens finder, step through the lens configurator, and read the selection logic condensed in telecentric lens selection: match magnification and image circle. Send part size, tolerance and nest variation with a quote request and the application team will check the sum — free expert support with every order. The rest of the lens guides sit on the lenses topic page.
Questions
How much does each millimetre of depth change cost on a standard lens?
The relative size error equals depth change divided by working distance. At 100 mm working distance, each millimetre of depth costs 1% of measured size; at 200 mm, 0.5%; at 400 mm, 0.25%. Multiply by the dimension for the absolute error — a 20 mm part gains 0.1 mm when it sits a millimetre nearer on the mid stand-off. It is entrance-pupil geometry, so no simulation is needed.
Does a telecentric lens also keep a depth-moving part in focus?
No. Magnification and focus are separate properties with separate depth bands. Magnification holds across the telecentric range, set by the chief-ray angle; focus holds across the depth of field, set by the f-number. A part can keep a perfectly constant size and still blur outside the depth of field. Design the nest into both bands, stop the iris down to stretch focus roughly in proportion to f-number, and budget the longer exposure or brighter light that costs.
What does a telecentricity value such as 0.1 degrees mean in micrometres?
It is the largest angle between the chief rays and the optical axis in object space. A point sitting Δs from the calibrated plane shifts sideways by about Δs × tan(angle): 1.7 µm per millimetre of depth at 0.1°, 8.7 µm at 0.5°. Compare that with your edge-detection repeatability, usually a fraction of a pixel, to decide whether the residual term belongs in the tolerance budget.
Why are telecentric lenses larger and longer than ordinary C-mount lenses?
The stop that makes chief rays parallel sits in front of the lens and must be at least as wide as the whole field, so barrel diameter grows with field of view rather than with sensor size. Length grows with working distance and magnification because the front and rear groups must place that stop at the focal plane. A 40 mm field needs front glass wider than 40 mm; there is no compact shortcut.
Can I just move the camera farther back instead of paying for telecentricity?
Partly. Relative error is depth change over distance, so doubling the working distance halves it — from 0.5% to 0.25% for the same nest slack. The image also halves, so the object-space pixel size doubles and small features lose resolution. Distance trades resolution for stability; it never makes magnification constant. Constant magnification under depth motion is exactly the telecentric case.

