Why a Telecentric Lens Gives a Double Image: Six Causes in Check Order

A second, offset copy of the part in the frame is one of the most misdiagnosed faults in machine vision, because the word "image" sends the investigation straight to the lens. In practice the optics are the last place to look. A double image is a second light path, and that path can be created by the illumination, the acquisition chain, any flat transparent element between the part and the sensor, or the mechanical stack holding the lens — all before a single lens element is implicated.

The order matters because the checks differ enormously in cost. Turning a light off costs seconds; rotating a protective window costs a minute; disassembling an optical stack costs an hour and risks introducing new faults. Working from the cheap end is the difference between finding the cause and disturbing a station that was running.

The observation that makes the checklist fast: does the second image move when the part moves? A real image moves and scales with the part. A ghost from a reflection inside the optical path shifts by a different amount, and may move the opposite way or stay fixed while the part translates.

Why "double image" is really a second light path

Everything on this list is the same event recorded twice: two exposures, two reflections, two sources, or two optical paths. The useful question is not what is wrong with the lens but where the second path branches off — at the source, in the air before the lens, at every glass surface in the assembly, at the sensor, or in the frame buffer.

Telecentricity makes one of these easier to identify. Because chief rays run parallel to the axis, a ghost from a tilted plane-parallel plate is displaced by a nearly constant amount across the field, unlike a defocused second image or motion blur.

The six causes, in the order worth checking

1. Two light sources, or one source firing twice

The cheapest test on the list: switch the illumination off one source at a time and watch the frame. A ring light plus a backlight, or two heads on one controller, will each produce a complete image of the part, and the offset between them is geometric rather than optical.

A single source can do it too when its strobe pulse overlaps two exposure windows, or when the controller fires per trigger and per internal timer. Freeze the trigger and run the source continuously; if the doubling disappears, it was temporal, not optical.

2. The acquisition chain, not the optics

Double exposure, a frame grabber delivering the same buffer twice, a trigger wired to both line-start and frame-start, or an encoder producing two pulses per position all produce two recordings of one image. The decisive test is a static part: if the doubling persists with the part still and the conveyor stopped, no motion is involved.

The second test is to capture a single frame with the longest possible exposure and no strobe. A genuine second light path persists; an acquisition artefact usually changes character or vanishes.

3. A plane-parallel element somewhere in the optical path

This is the most common optics-adjacent cause, and it is still not a lens fault. Any flat, parallel, partially reflecting surface between the part and the sensor produces a laterally displaced secondary reflection: the sensor cover glass, a camera protective window, an IR-cut filter, a glass guard plate, or a transparent fixture plate.

The practical test does not need the formula: rotate the suspected plate. If the second image swings around the field as the plate rotates, the plate is the cause. If nothing moves, it is not.

Removing the plate is the confirmation. Where a protective window is genuinely required, a wedged window rather than a parallel one removes the ghost path.

4. A coaxial illumination path inside the lens

Coaxial telecentric lenses contain a beam-splitter so illumination can be delivered along the imaging axis. That is deliberate and useful on specular parts, and it is also a surface that can produce a faint second image if the port is fed at an angle or with a source that was never collimated to it.

POMEAS publishes coaxial variants explicitly: PMS-LTC2-110C (2.0×, 110 mm working distance, 8 µm resolution, TV distortion 0.01%, depth of field 0.5 mm, built-in coaxial split path, maximum 2/3-inch sensor) and PMS-LTC1-110C (1.0×, 110 mm, 13.42 µm, TV distortion 0.02%, depth of field 1.6 mm at 1:1, maximum 1/2-inch sensor, built-in coaxial split path).

The test is to disconnect the coaxial light source and re-capture. If the doubling disappears, the cause is the illumination coupled into the port, not the imaging optics.

5. The mechanical stack: spacer, adapter, or extension tube

Anything added between the lens and the camera changes where the image plane sits and can add a reflecting surface. C-mount spacers, filter holders, adapter rings and extension tubes are all candidates, often left by an earlier integration step with no record.

The test is to strip the stack to the bare lens and camera, capture, then add components back one at a time. Because the working distance of a telecentric lens is fixed, if the bare combination focuses at the published distance, the stack was what changed it.

6. Defocus and vibration mistaken for doubling

A shallow depth of field combined with part motion can produce an edge that looks doubled when it is actually smeared. PMS-LTC2-110 publishes a depth of field of 0.5 mm at 2.0×, and PMS-LTC1-85 publishes 1.00 mm at 1.0× with 85 mm working distance and 8.39 µm resolution. At those depths of field, vibration of a fraction of a millimetre is enough to produce a second-looking edge.

The distinguishing test is directional: motion blur is aligned with the motion, a ghost is not.

Where POMEAS telecentric lenses land on those six

Published values for the models most often involved in a double-image investigation, including the coaxial variants:

Reading the table against the six causes:

- PMS-LTC2-110C and PMS-LTC2-110 share magnification, working distance, resolution and depth of field, and differ in the optical path: the C variant carries the built-in coaxial split. On a C variant, cause 4 applies; on the non-C variant the same symptom must come from one of the other five. - PMS-LTC1-110C is the coaxial option at 1.0× and 110 mm, published with 13.42 µm resolution and 1.6 mm depth of field at 1:1 — the largest here, which makes pure defocus the least likely explanation among these five. - PMS-LTC1-85 publishes the lowest distortion in this group at 0.002% TV and the shortest working distance at 85 mm. Neither bears on ghosting; both are listed because a station diagnosing a double image often has to re-derive its own setup numbers at the same time. - PMS-LTCM2-65-S publishes 2.2 µm resolution and NA 0.152 with centre MTF greater than 50% at 100 lp/mm, but no depth of field and no distortion. Both are "not published", and no comparison should be drawn from them.

The five models are listed with the fields each one publishes. A blank is a gap in published data, not a value that can be carried across from a neighbouring row.

What those numbers mean on the line

Scenario 1 — Doubling that vanishes when the second light is unplugged

A station images a polished metal bracket with a ring light and a separate coaxial source, and two complete outlines appear at a fixed offset. Turning off the ring light removes one. This is cause 1, and no lens specification is involved.

Scenario 2 — Doubling present with the part held still and the line stopped

The part is clamped, the conveyor is off, and the frame still shows two outlines at a fixed offset. This eliminates motion and most mechanical causes at once and points at the acquisition chain (cause 2) or a fixed reflecting surface (cause 3). Capturing with the strobe disabled separates them.

Scenario 3 — A ghost that swings when the protective window is rotated

The station runs behind a flat glass guard plate, and rotating the plate moves the second image around the field. The plate is the cause. The fix is mechanical: replace the parallel plate with a wedged one, or tilt it far enough that the reflection leaves the field.

Scenario 4 — A "second edge" that disappears when the conveyor is stopped

A station running PMS-LTC2-110 at 2.0× with a published depth of field of 0.5 mm reports a doubled edge on a moving line; stopping the conveyor produces a single sharp edge. This is not a ghost at all but motion within half a millimetre, and the remedies are exposure time, strobe duration and damping.

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 causing a double image on a specific installation, and none should be inferred. Identifying the branch point of the second light path requires access to the station, and the ordering here is about cost and probability, not certainty.

Optical performance is not measurement accuracy, and a clean image is not a correct measurement. Removing a ghost improves the image; whether it improves the measurement depends on which edge the algorithm was locking onto.

Every figure here is the published value for the specific POMEAS model named beside it, not measured under a single common condition. Where a field is absent — depth of field and distortion for PMS-LTCM2-65-S — read it as "not published".

The telecentric content already on this site treats magnification families, sensor matching and working-distance selection separately. This article addresses a symptom after a station is running; selection questions belong to those pages.

FAQ

Is a double image usually a lens defect?

Rarely. The most frequent causes are a second light source, a source firing within two exposure windows, an acquisition artefact, or a flat transparent plate somewhere in the optical path reflecting a second time.

How do I tell a ghost from motion blur?

A ghost moves differently from the part and stays present with the part clamped. Motion blur is aligned with the direction of travel and disappears when the conveyor is stopped.

Do coaxial telecentric lenses cause ghosting?

The built-in beam-splitter in a coaxial model such as PMS-LTC2-110C or PMS-LTC1-110C is a deliberate surface in the path. Disconnecting the coaxial source and re-capturing separates the illumination path from the imaging optics.

Why does the second image not scale with the part?

A ghost produced by a fixed reflecting surface is displaced by a fixed geometric amount set by the plate thickness and the ray angle, so it does not track the part the way a real image does.

Does depth of field affect whether I see doubling?

Yes, indirectly. PMS-LTC2-110 publishes 0.5 mm of depth of field at 2.0×, so vibration or motion within that range produces an edge that reads as doubled when it is actually smeared.

What should I record before contacting the manufacturer?

The model, the actual working distance, the sensor format, the illumination geometry, whether the doubling persists with the part stationary, and whether it changes when the strobe is disabled.

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.