Telecentric vs FA Lens in Precision Measurement: 4 Checks

This article is also available in 简体中文.

Short answer: in precision measurement the difference is not that "telecentric is more accurate". A telecentric lens collapses three independent error mechanisms — edge-position distortion, magnification drift with part height, and scale change on refocus — into one bounded residual, while a standard FA lens lets all three scale with your fixture. Four checks decide it: field of view against magnification, optical resolution against pixel pitch, depth of field against the real spread of part heights, and the residual error budget from telecentricity and TV distortion.

This update keeps the optical reasoning and cost logic of the original comparison and adds the layer that selection actually turns on: what happens on real parts — height steps, free edges, through-holes measured on one axis — and when a well-calibrated FA lens is the better engineering answer.

Part of the POMEAS Machine Vision Optics Comparisons series — see the Machine Vision Optics Comparisons index for all 15 articles in this set, grouped by the option being compared.

When the comparison lands in favour of telecentric, the 1.5× group is the fastest way to see whether a 1.5× model fits your field and space.

Why geometry, not tolerance, sets the ceiling

In a telecentric design the aperture stop sits at the rear focal plane of the front group, so the chief ray at every field point leaves parallel to the optical axis: orthographic imaging. In a standard FA lens the chief rays converge toward the stop: perspective imaging. Polishing cannot change that, and calibration removes only a static warp. Telecentricity, quoted as a maximum ray angle, is what ties optics to mechanics, because it converts Z-direction part movement into lateral measurement error. POMEAS publishes object-side telecentricity below 0.02° for the 1.2× and 1.5× coaxial telecentric lenses, below 0.016° object-side and below 0.069° image-side for the 2.0× LTCM2-65C-S, and below 0.05–0.1° object-side for the 1.0× bi-telecentric design for 1.1″ sensors.

Telecentricity and distortion on real parts

A distortion percentage is a field average. On a real part, what matters is which feature you measure, because each feature excites a different error term.

Two of these three cases are invisible in a distortion figure. A lens can hold 0.01% TV distortion and still fail on a stepped part if its telecentricity is loose, and it can be tightly telecentric and still fail a two-sided measurement if only the object side is corrected. That is why the 2.0× LTCM2-65C-S publishes both figures, and why the 1.1″ bi-telecentric design publishes object-side telecentricity alongside image-side distortion.

Depth of field when part heights differ

Telecentricity keeps magnification stable with object distance; it does not keep the image sharp. For a part whose height varies, depth of field becomes the limiting specification — and it falls steeply with magnification.

Sensor formats differ: the WD110 series is designed for 2/3″ sensors, LTC1-200C for 1/1.8″. All values are published typicals for the stated configuration; confirm against the current selection manual. The rule that follows is blunt: the full stack-up of thickness tolerance, fixture wear and conveyance repeatability must sit inside the depth of field, not on its edge. Dropping from 1.2× to 0.5× multiplies usable height tolerance roughly fourfold and costs resolution — 6.71 µm becomes 12.43 µm.

Measurement repeatability: the factor checklist

Repeatability is what the customer finally measures: same part, same station, same number. Eight factors move it, and only the first four are optical.

When an FA lens is the better answer

Over-specifying optics costs as much as under-specifying them, and a well-calibrated FA lens with good illumination carries a very large share of vision work. Speed is the row most often forgotten: a telecentric design runs at a small aperture and needs more light or a longer exposure on the same part.

Two structural constraints sit behind that table: a telecentric front group must be at least as large as the field of view, and its magnification is guaranteed only at the design working distance. Standard FA distortion, magnification sensitivity with object distance and depth of field are all model-specific — confirm them for the exact lens you intend to use (to be confirmed). Do not carry a catalogue-class figure into a tolerance budget; our note on when a telecentric lens is the wrong choice covers the failure modes.

Four checks before you specify

Step 1 — Fix the field of view, then derive the magnification. β = sensor dimension ÷ field dimension. For a 7.3 mm required field width on a 2/3″ sensor (8.8 mm wide), β = 8.8 ÷ 7.3 ≈ 1.2, so the 1.2× LTCM12-110C coaxial telecentric lens is the first candidate: published 2/3″ field 7.3 × 5.5 mm at 110 mm working distance.

Step 2 — Check resolution against pixel pitch. Object resolution 6.71 µm at 1.2× becomes 6.71 × 1.2 ≈ 8.05 µm image-side (calculated from the magnification relation). Under the common rule that a lens should resolve no coarser than twice the pixel pitch, the pitch must be about 4.0 µm or larger. With smaller pixels, the 1.5× LTCM15-110C is the tighter option at 5.59 µm.

Step 3 — Check depth of field against the real height spread. LTCM12-110C specifies 0.67 mm. If thickness tolerance plus fixture wear plus conveyance repeatability exceeds that, step down the magnification ladder — 0.5× gives 2.96 mm at 12.43 µm — or measure face by face. Never plan a stepped part to sit exactly at the depth-of-field edge; the trade-off is worked through in how depth of field is improved on industrial lenses.

Step 4 — Close the error budget at the field radius you use. Two residual terms remain. Distortion error is roughly field radius × TV distortion: at a 3.0 mm radius with 0.01%, about 0.3 µm (calculated). Telecentricity error is roughly Z drift × tan(telecentricity): with <0.02° and 0.2 mm of fixture movement, about 0.07 µm (calculated). Both leave room inside a ±10 µm tolerance, which is the point of the exercise. The full workflow is in the telecentric lens selection guide.

Where to start from the sensor end

For 2/3″ sensors at 110 mm working distance the coaxial series runs 0.3× to 2.0×. For 1.1″ sensors the 1.0× PMS-10VP110C111T bi-telecentric lens gives a Φ18.40 mm object field, and the 0.37× LTCM037-150-P is also a 1.1″ design. For tight telecentricity at long working distance, the 1.0× LTC1-200C at 200 mm working distance is the reference point. The full range is listed in the telecentric lens product family.

If you are sizing a station now, send us the sensor format, the field of view and the feature tolerance. We will return a magnification, working distance and depth-of-field combination and tell you where the four checks land. Contact POMEAS with your drawing or measurement requirement; for background on the application, see telecentric lenses for dimensional measurement and inspection.

Frequently Asked Questions

Can a telecentric lens measure a part with a 2 mm height step?

Only if 2 mm fits inside depth of field. On the WD110 coaxial series that means 0.2× (10 mm), 0.3× (4.44 mm) or 0.5× (2.96 mm). At 1.2× the published figure is 0.67 mm, at 1.5× 0.44 mm and at 2.0× about 0.27 mm. If a 2 mm step has to be measured together with 5–7 µm resolution, the practical answer is usually to acquire the two faces in separate images, or to accept a lower magnification for the height-tolerant measurement and a higher one for the fine feature.

Does a telecentric lens remove the need for calibration?

No. Telecentric optics remove the need to recalibrate when part height changes, because magnification is independent of object distance. They do not remove the need to establish the pixel-to-millimetre scale, and they do not make distortion zero: published TV distortion on the WD110 coaxial series is 0.01–0.03%, and the 1.1″ bi-telecentric model is specified at image-side distortion below 0.05–0.1%. Recalibrate after any change in optics, illumination or fixturing.

Why does same-axis two-sided measurement need bi-telecentric optics?

Because two independent scales must stay fixed. The object side must hold its scale as the measured face changes height, and the image side must hold its scale if focus or sensor position moves. LTCM2-65C-S publishes both figures: object-side telecentricity below 0.016° and image-side below 0.069°. The 1.0× bi-telecentric lens for 1.1″ sensors publishes object-side telecentricity below 0.05–0.1° and image-side distortion below 0.05–0.1%. A purely object-side telecentric lens does not guarantee the second condition.

What actually limits measurement repeatability?

In practice, in this order: illumination stability, fixture and conveyance repeatability, depth-of-field margin, and the sub-pixel edge algorithm. Telecentricity and residual TV distortion set the floor of achievable error — below 0.02° and 0.01–0.03% on the WD110 coaxial series — but on most stations the mechanical and lighting terms are larger. Tighten those before paying for a tighter optical grade, and use coaxial illumination on specular parts.

When is an FA lens the better choice for measurement?

When the feature tolerance is loose relative to the field of view, when the field of view exceeds any mountable barrel, when working distance must stay adjustable, when only one face carries the dimension and part height is controlled, or when cycle rate and cost dominate. Standard FA lens distortion, magnification sensitivity with object distance and depth of field are all model-specific; confirm them for the exact lens before putting them in a tolerance budget (to be confirmed).

Which POMEAS lens should I start from?

Work from the sensor and the field. For 2/3″ sensors at 110 mm working distance the coaxial telecentric series covers 0.3× to 2.0× (11.18 µm to 6.71 µm object resolution, 4.44 mm to 0.67 mm depth of field). For 1.1″ sensors the 1.0× bi-telecentric model gives a Φ18.40 mm object field, and LTCM037-150-P is also a 1.1″ design. For long working distance with tight telecentricity, LTC1-200C is specified at 200 mm. All figures are published typicals for the stated configuration.

Related POMEAS product categories

The hardware referred to above is grouped into the following categories on the product side, with the specifications and model list for each range.

• Telecentric Lenses — object-side and bi-telecentric lenses for distortion-free dimensional measurement.

FA Lens Distortion Figures Across the Range

Measured spread

The comparison above is about principle. This section adds the measured figures, because the decision usually turns on how much distortion a specific model carries rather than on whether a standard FA lens has any. Across the POMEAS FA models that publish both figures, focal length spans 8 to 82 mm (n=30) and TV distortion spans 0.002 to 3.73 % (n=30). That spread is wide enough that the family average is not a useful number. A selection that assumes a typical FA distortion figure will overestimate some models badly and underestimate others.

Wide angles carry most of it

The short focal length models account for the upper end of that range. A comparison between a telecentric lens and a wide-angle FA lens is not a fair comparison against the whole FA family, because most FA models sit well below the worst case. This matters when the comparison is used to justify a purchase. Presenting the worst FA figure next to a telecentric figure makes the gap look larger than it is for the specific model that would otherwise have been chosen.

Field position matters

Distortion grows toward the field edge, so a part measured at the centre of one frame and at the edge of another is measured under two different error conditions. A telecentric lens does not eliminate this, but its distortion range is 0.002 to 0.06 % (n=31), which is far below the FA spread. For a measurement application the practical question is the distortion over the region actually used, not the figure over the whole frame. Restricting measurement to the central field is often enough to close the gap without changing the lens.

Where the cost difference comes from

The price gap between a telecentric lens and a standard FA lens of similar focal length is large, and it buys two things: the perspective-error immunity that comes from accepting only near-parallel rays, and the tighter distortion figures quoted above. If the application measures a dimension on a part presented at a controlled height, both properties are used and the price is justified. If the application only detects presence or classifies, neither property changes the result and the standard lens is the correct choice. Recording which of the two properties the application actually needs makes the purchase decision checkable rather than a matter of preference.

Measured Distortion Across the Telecentric Range

The Figures Behind the Comparison

The comparison above explains the mechanism. This section gives the measured figures so the argument can be checked against actual models rather than accepted as a principle.

TV distortion across the POMEAS telecentric models spans 0.002 to 0.06 %. That is a span, not a single value, and the low end of it is the figure usually quoted when a telecentric lens is compared with a standard FA lens.

Telecentricity Is a Separate Figure

Distortion describes where a point lands in the image relative to where it should land. Telecentricity describes how parallel the chief rays are, and therefore how much the image moves when the part moves in height.

Across the POMEAS telecentric models that quote it, telecentricity spans 0.02 to 0.04 deg. It is quoted on fewer models than distortion, which is worth noting when a specification depends on it: a model that does not quote the figure has not been characterised, which is not the same as being good.

A Tolerance Stack Needs Both

A distortion figure alone does not predict how a measured dimension changes when the part moves in height. That change is the telecentricity contribution, and on a standard FA lens it is the dominant error for any part presented without a hard stop.

Writing the stack explicitly, with a term for distortion and a term for height-induced magnification change, shows which of the two the application actually depends on. Applications that present the part against a stop can often live with a lens that is weaker on telecentricity but tighter on distortion, and that is usually the cheaper lens.

Why the Span Matters More Than the Average

A family average distortion figure is not useful for a purchase decision, because the models at the low end of the span are available at similar prices to the ones at the high end.

The practical consequence is that the figure has to be read per model, at the working distance and magnification that the application uses. A quotation that cites a family figure without naming the model leaves the buyer unable to check the claim.