Laser line profile with telecentric lens for presence checks

Retaining rings, pins, clips and circlips go in wrong or not at all, and a resistance probe reads a half-seated ring as good until wear turns it into a permanent false pass. There is an optical route that removes both failure modes: project a Class 1 line laser across the seat, photograph the bright line it draws with a telecentric lens, and pattern-match the profile against three taught states — ring seated, ring missing, ring partly out. Assembly play cannot change the image scale, hall lighting cannot outshine a filtered laser line, and the vision system hands one state code to the PLC. What follows is the physics that makes the method robust, a fully sized worked example at 0.5x, the magnification ladder for other part sizes, and the pitfalls that cost integrators weeks.

How the check works

A line projector sits above the station, roughly over the part's final ring position, and sweeps a thin fan across the groove and the land beside it. The camera hangs above that, looking straight down through an object-side telecentric lens. Where the line crosses flat land it runs straight; where it crosses a ring edge it steps sideways and kinks; where the ring is missing it runs on through the open groove. Those three shapes are the whole signature.

The vision job scores the captured line against the three taught references, takes the best match, and writes a state code to the PLC. No brightness threshold and no reflection model: the geometry of one bright line does the talking, which is why the same station works on bright, dark and oiled parts.

Why the lens must be telecentric

Pistons float in their fixtures. The bore centre wanders laterally by a millimetre and the part sits high or low with the skirt batch. An ordinary lens turns that wander into changing magnification:

At a 110 mm working distance, 1 mm of axial play swings the scale by about 0.9%, which drags a profile feature sitting 5 mm from the image centre by roughly 45 µm — the same order as the seating error you are chasing. The taught references go stale within a shift.

An object-side telecentric lens accepts only rays running nearly parallel to the axis, so magnification holds constant across the depth range and across the field. The profile of a floating part stays congruent with the taught profile, and the match score measures the ring, not the fixture. The trade is size: telecentrics are physically longer and need more working distance, and the field is fixed by the design. The entocentric versus telecentric comparison and the piece on when a telecentric lens earns its cost for gauging cover when that trade pays.

Telecentricity carries its own error budget; the telecentric distortion and telecentricity explainer shows how to read it off a data sheet.

Why a line laser beats a reflection check and a probe

Reflection-based optical checks photograph the ring under white light and judge what comes back. Machined steel punishes that approach: face brightness changes with finish lot by lot, and a shaft of sunlight through a roof light flips the verdict. A mechanical probe dodges neither problem and adds its own — it wears, it pushes on a fragile ring, and it cannot reliably tell a ring seated proud from one seated flush.

The line laser sidesteps all three. Its output is monochromatic and, inside its narrow band, orders of magnitude brighter than hall lighting, so a bandpass filter in front of the lens passes the line and rejects almost everything else. What the camera sees is one bright line on a near-black field whatever the workshop lights are doing. Because the match keys on the line's shape, finish-driven changes in absolute brightness barely register. And because nothing touches the part, nothing wears out.

Worked example: sizing the optics for a 12 mm ring profile

The example part is a small-engine-class piston whose ring region — groove plus one land either side — needs a 12 mm × 9 mm field, and the fixture lets the part wander ±1 mm axially. Those two are the application inputs. Everything else comes off the data sheets of the two parts below and the arithmetic between them.

Two catalog parts carry the example, camera first:

Then the lens:

The rule to apply, in words: field of view equals sensor size divided by magnification, and object-space pixel equals pixel pitch divided by magnification. Both are telecentric staples — the field of view from magnification and object space pixel size walkthroughs carry the derivations.

A 1/1.8" sensor measures about 7.2 mm × 5.4 mm with a 9 mm diagonal, and 6 MP at the usual 3:2 layout is about 3072 × 2048 pixels, so the pixel pitch is 7.2 mm ÷ 3072 ≈ 2.3 µm. The lens's 1" image circle covers that 9 mm diagonal with room to spare.

Field of view: 7.2 mm ÷ 0.5 = 14.4 mm wide, 5.4 mm ÷ 0.5 = 10.8 mm tall. The 12 mm × 9 mm region fits with 1.2 mm spare per side horizontally and 0.9 mm vertically — margin that absorbs fixture drift without a re-teach.

Object-space pixel: 2.3 µm ÷ 0.5 = 4.6 µm. A ring edge of about 0.5 mm spans 0.5 mm ÷ 4.6 µm ≈ 109 pixels, and a seating offset of 0.05 mm displaces the line by 0.05 mm ÷ 4.6 µm ≈ 11 pixels.

The height kink comes from the projection angle. With the fan tilted 30° off the camera axis and the ring proud of its seat by 0.3 mm, the line steps sideways by 0.3 mm × tan 30° ≈ 0.17 mm, which reads as roughly 38 pixels of jump — impossible to confuse with seated.

Decision: this pair resolves seating displacement in the tens of micrometres with double-digit pixel margins, and ±1 mm of play fits the depth of field this class reaches stopped down, as the next section quantifies. The rule that falls out: fit a 0.5x-class telecentric with 100-120 mm of working distance for regions of 10-16 mm; step up in magnification below 7 mm of region and down toward 0.17x above 20 mm. Run your own region through the camera resolution calculator to check the pixel margins before ordering:

Choosing the magnification class

Three catalog lenses bracket the useful range, all C-mount object-side telecentrics:

All 3 share: Mount C-mount. Values as printed in each product's data sheet; open a part for its drawing and full specifications.

On the example camera those magnifications give fields of about 5.3 mm (7.2 mm ÷ 1.36), 14.4 mm and 43 mm across. Take the middle class for a region near 12 mm, as the worked example does. The coarse class still separates states at a 43 mm field — its 13.9 µm object pixel gives a 0.05 mm offset roughly 4 pixels — but that is classification, not measurement; a dimensional read wants a smaller field or finer pixels. Match telecentric magnification to image circle for the sensor you pick, then let the telecentric lens finder sort candidates by both.

Geometry: angle, depth of field and the filter

The projection angle converts height into visible sideways shift:

θ is the angle between fan and camera axis. Work between 20° and 35°: the kink stays measurable without the lens barrel shadowing the groove or the line tearing on steep walls. Below about 10° the height information thins — the same 0.3 mm step yields only 0.05 mm of shift, some 11 pixels at the example's 4.6 µm object pixel, workable but thin. Above 45° you trade kink depth for shadowing and flare.

Depth of field must swallow the play band. At 0.5x, stopping the iris down two or three stops typically reaches multi-millimetre depth; for a ±1 mm band, ask for at least 3 mm of depth of field and keep the play centred inside it. The telecentric depth-of-field validation shows the drift check that goes with the claim, and the depth-of-field calculator runs your own numbers.

Put a bandpass filter matched to the laser wavelength at the front of the lens — 650 nm red and 520 nm green are the common line-laser bands. Set exposure so the line's peak sits near 80-90% of full scale while the background stays under 10%, and strobe the projector so it emits only during exposure. Hall lighting, skylight and stray glint then stop existing as far as the sensor is concerned. Placement and mounting details live in the telecentric lens filters guide.

Teaching the three states and handing the result to the PLC

Teach from golden parts, not simulation: one ring seated flush, one with the ring removed, one seated proud by the largest offset you intend to reject. Capture ten to twenty images per state across production batches, average each set into a reference profile, and store per-point tolerance bands with it. Score the live line point by point — a single aggregate can pass while one end of the line sits a groove-width high, the same reason metrology profile checks compare point-level deviations instead of one overall number.

The PLC conversation stays deliberately boring. The cell sends the variant ID with the trigger; the vision job loads that variant's reference set, matches, and returns a state code — 0 seated, 1 missing, 2 mis-seated — with a pass bit and the score. Manual stations run the same job behind an operator button; automatic ones trigger off the part-present sensor. New variants bring new reference sets, never new optics, as long as their regions fit the field.

When this method is the wrong tool

State matching answers one question — is the ring where it belongs? Reach for something else when the question changes.

• You need a dimension, not a state: a gap width held to ±0.05 mm wants a calibrated 2D or 3D laser profiler that returns numbers you can chart.
• The part never stops. A stopped station wants stillness; for continuous web or conveyor work the line scan illumination techniques with encoder-synced exposure are the built-for-it path.
• The ring hides inside a bore with no line of sight. No fan reaches it; a probe or borescope stays the honest answer.
• The ring is a bright curved spring whose line scatters into noise. Try coaxial illumination through a telecentric with a light port — the in-line illumination guide covers when that wins — before abandoning optics.

One related limit: an LED line scan light behind a slit can draw a bright line, but its broad spectrum cannot be filtered the way a laser's narrow band can, so ambient immunity suffers. For a hall-light-proof presence check, keep the laser as the source.

Checks before you order

• Field: region width plus at least 25% margin sets the magnification class — 12 mm of region put the example at 0.5x.
• Depth of field: printed depth at your f-number must cover the play band with the play centred; 3 mm or more for ±1 mm at this magnification class.
• Image circle over sensor diagonal, and the mounts matched — C-mount throughout the example.
• Bandpass ordered together with the projector, wavelength-matched and with clear aperture covering the lens front.
• Mono sensor: a colour array spends most of its photosites away from the laser band and throws away signal.
• Class 1 projector with a strobe input, so it emits only during exposure.
• One mapping table for the state codes, shared by the HMI, the PLC program and the vision log.

Fit it to your region

Scale the example instead of restarting: divide your region width into your sensor width and the magnification class falls out; divide the pixel pitch by that magnification for the object pixel and the pixel margin on your smallest seating error. When the numbers land between classes, take the finer one and keep the spare field — alignment margin costs less than a re-teach. The telecentric lens category and the area scan camera category hold the candidates, and the lens configurator walks the pairing. If the golden-part set or the PLC mapping needs a second pair of eyes, contact us.