Color Mark Sensor for Packaging Machines: How It Works and How to Choose
A color mark sensor turns a printed eye mark into a position reference for cutting, sealing or labeling. Choose it for the finished film, the mark and light-spot geometry, and the controller’s ability to capture the signal. A fast sensor alone does not guarantee an accurate cut.
What does a color mark sensor do on a packaging machine?
It detects a repeatable optical change between a printed registration mark and its surroundings. The machine uses that event to relate the printed design to its next operation. The sensor finds the reference; the controller and mechanism determine what happens next.
On a flow wrapper, the reference helps align the artwork with the seal and cut. On a vertical form-fill-seal machine, it can support film-feed correction. On a label or converting line, it can reference a cut, print or placement operation. These are different control tasks, not different guarantees supplied by the sensor.
Contrast detection is not the same as color identification
“Color mark,” “eye mark” and “registration mark” often describe the packaging application. A contrast sensor compares reflected-light levels against a threshold. A color sensor evaluates whether the target matches a taught color within its configured tolerance. Some devices offer both modes.
Start with contrast sensing when the task is one reliable mark edge per package repeat. Consider color discrimination when similar-brightness artwork must be distinguished by color, and vision or pattern sensing when the reference is a symbol, shape or printed feature rather than a simple mark. A label gap is a different target from a printed eye mark.
What RGB actually changes
Red, green and blue illumination can produce different contrast from the same ink and background. Some sensors select the most useful emitter during teach-in; other designs use white light and evaluate reflected color channels. “RGB” alone does not tell you which method or output data the device provides.
Judge the separation between the mark and every unwanted condition in the active sensing mode. A color combination that looks obvious to a person is not automatically easy for the sensor, especially after coating or lamination.
Which mark layout and light spot will work reliably?
Choose a clear sensing path through the artwork, then match the mark dimensions to the spot and web movement. A high-contrast mark can still fail if another printed feature produces an indistinguishable signal along the same path.
Check the whole repeat, not just the eye mark
Follow the line of print that passes under the stationary sensor through one complete package repeat. Include text, barcodes, borders and seams. Reserve a clear mark lane where possible; otherwise test the strongest competing feature as well as the intended mark.
If two features cross the same threshold, changing sensitivity may simply exchange one false trigger for one missed mark. Changing the scan lane, the mark construction or the sensing method can be the more useful next trial. SICK’s registration-mark design guidance explains why surrounding artwork belongs in this decision.
Check dimensions in both directions
The mark dimension along web travel affects how long it passes the spot. Its dimension across the web must accommodate the spot and lateral tracking variation. For a rectangular spot, its orientation matters: rotating the sensor can exchange the long and short dimensions relative to motion.
Use the spot dimensions at the actual working distance, not the housing size or a laser-looking aiming point. Aim for distinguishable mark and background signals throughout the allowed movement. The switching edge can occur while the spot partly overlaps the mark; a geometric mark-size rule does not predict the exact electrical pulse.
How do glossy, metallic and clear films change the choice?
They change the light reaching the receiver, sometimes more than the ink does. Test the finished substrate, print, coating and backing together. A successful paper proof is not evidence that the laminated production web will behave the same way.
| Material | What can change | What to compare |
|---|---|---|
| Matte paper or film | Ink density and competing artwork reduce mark separation. | The weakest intended mark and strongest unwanted feature, using the same settings. |
| Glossy laminate or metallized film | Glare, wrinkles and web angle alter the received signal; a strong return can saturate the receiver. | Final laminate at the expected distance and angle extremes, including roll-to-roll variation. |
| Clear film | The backing or product behind the film contributes to the optical return. | The printed mark with the actual machine backing and product conditions. Consider a controlled backing or opaque patch where appropriate. |
Look for useful signal or contrast diagnostics, suitable distance tolerance and an illumination method that separates the real samples. These features help evaluate a setup; none is a blanket guarantee for reflective packaging.
Do not add an arbitrary tilt to every sensor. For example, the Banner R58E manual specifies a 10 mm ±3 mm mounting distance and approximately 15° off perpendicular for reflective material. Those are R58E instructions, not universal settings for color mark sensors. A different optical design may require different geometry.
How should you mount and teach the sensor?
Establish the final optical geometry before teaching, and verify the resulting output before adjusting registration offsets. Teaching a hand-held sample and then changing the distance or backing changes the conditions the sensor learned.
- Make the installation safe. Follow the machine’s isolation procedure before changing mounts or wiring. Carry out powered trials only under the approved commissioning procedure and guarding.
- Secure the reading position. Use the specified distance and orientation, a rigid bracket, and a web location with controlled flutter. Keep the spot in the intended lane throughout normal movement.
- Teach representative samples. Use the finished mark and background. Static teach presents conditions separately; dynamic teach samples moving material where the model supports it. Follow the exact sequence and confirm a successful teach indication.
- Test every competing condition. Check the complete print repeat and expected material variation. Record the selected mode, source color and signal separation if the sensor exposes them.
- Verify the receiver sees the intended event. Confirm the electrical output, active state, chosen edge and input capture before changing the machine’s cut or seal offset.
NPN/PNP describes the electrical output circuit; light-ON/dark-ON describes optical-state logic. Color-match modes may instead use match/mismatch logic. None of these names establishes the connector pin assignment, PLC common or required input timing.
How fast must the sensor and controller be?
Check the shortest usable signal at maximum web speed, the selected sensor mode and the receiving input—not only marks per minute. Then separate repeatable delay from timing variation when estimating registration error.
A low mark rate can still produce a short pulse
Illustrative calculation. Assume a web runs at 120 m/min, with a 4 mm mark dimension along travel and a 200 mm repeat pitch.
- Web speed: 120 ÷ 60 = 2 mm/ms.
- Geometric mark-passage time: 4 ÷ 2 = 2 ms.
- Repeat period: 200 ÷ 2 = 100 ms, or only 10 marks/s.
The machine has plenty of time between marks, but the mark itself passes quickly. The actual output pulse may differ from 2 ms because of spot overlap, threshold, response and timers. Measure its shortest active and inactive intervals at the receiver and compare them with the input’s documented requirements.
A short pulse can occur entirely between ordinary PLC samples. Where supported, a hardware latch or registration input can capture an event independently of the cyclic program. It still needs compatible wiring, valid pulse widths, correct arming and a suitable position reference.
The operating mode changes the timing specification
Documented model example: OMRON lists 50 µs operate/reset response for the E3NX-CA11 in super-high-speed mode, only in Contrast Mode. The listed high-speed and standard modes are 250 µs and 1 ms respectively. Its Color Mode uses RGB-ratio discrimination. Choosing the same amplifier does not make the fastest timing apply to every configuration. See the E3NX-CA specifications; these are not xsz sensor ratings.
Delay travel is not the same as cut-position scatter
At the example speed of 2 mm/ms, an assumed constant 0.6 ms event delay corresponds to 1.2 mm of web travel. A separate assumed 0.1 ms range of timing variation corresponds to 0.2 mm of position variation at constant speed.
The first may be compensated by an appropriate control strategy; a single offset does not remove the second. Neither calculation is a finished-cut tolerance. Mark-edge quality, capture timing, encoder scaling, film stretch or slip, and mechanism repeatability remain part of the machine result. During acceleration, the constant-speed calculation is only an approximation.
Why does the machine miss marks, double-count or cut off position?
First locate the disagreement: optical detection, electrical pulse, controller event or finished-package position. Keep the known settings fixed while comparing these stages; changing all of them together destroys the evidence.
Two edges do not necessarily mean two detections
In a simple untimed setup configured ON for the mark, one mark normally produces one ON interval: a rising edge on entry and a falling edge on exit. If the controller counts both edges, one optical event can become two counts without sensor chatter.
Several separate ON intervals within one intended mark are different: investigate changing reflection, print detail, alignment and electrical integrity. Pulse stretching may help a compatible input capture a short pulse, but excessive stretching can remove the inactive gap between marks; it is not a general cure for false detections.
| Observation | Next discriminating check |
|---|---|
| Mark/background signal separation disappears | Compare old and new film under unchanged geometry and settings. Check laminate, glare, backing, distance and taught mode before compensating a controller offset. |
| Extra electrical pulses appear | Correlate each pulse with a location in the complete print repeat. Separate a competing graphic from chatter within the mark; also check signal integrity. |
| A clean pulse reaches the input, but a capture is missing | Compare measured pulse widths and voltage levels with the input configuration, filter, selected edge and latch rearm requirements. A flashing sensor LED alone cannot establish this observation. |
| Captures match, but cuts drift or scatter | Compare captured encoder positions with actual cut locations. Check encoder reference, phase correction, web slip/stretch and mechanics while keeping the optical baseline unchanged. |
Illustrative troubleshooting scenario: suppose a new print roll produces an extra pulse at the same point in every repeat. If an added graphic crosses the scan lane there, that observation favors an artwork-discrimination problem over a sensor-speed problem. Trial a clear lane or a more suitable discrimination mode. Only a new waveform and capture comparison can confirm whether the change worked.
A controller acceptance window can reject marks outside an expected position range where the control architecture supports it. It cannot recreate an unreadable mark, and the commissioning plan must define window initialization, restart and what happens when the expected mark is missing.
What should the supplier recommendation and production trial prove?
They should connect a specific sensor configuration to reliable detection of your finished packaging—not merely show that a sample LED can switch. Send a real film sample and enough machine information to make that connection testable.
- The optical task is reproducible. Provide the full repeat artwork, mark dimensions, travel direction, final laminate or backing, and expected material variation.
- The proposed installation is feasible. Record the working distance, spot orientation, web-position tolerance, mounting space and relevant cleaning or environmental exposure.
- The timing and interface are compatible. State the speed range and required registration tolerance. Match the exact sensor, mode, response, output logic, connector and timer settings to the actual controller input.
- The whole event chain is checked. Compare intended marks, sensor pulses, accepted controller events and finished cut/seal/label positions. Define how splices, missing marks and intentional ignored events are treated.
- The result survives normal changeover. Include startup, acceleration, full-speed running, representative roll lots and the allowed web movement. Record the trial quantity, observed misses/extra events, position variation and all settings.
Set the test duration and acceptance criteria for the process; there is no universal sample count that proves every packaging line. Keep the accepted film reference, exact model and configuration, bracket position and controller recipe together so the next changeover starts from a known baseline.
For a model discussion, the xsz sensor color mark sensor range is a starting point. Ask for the exact variant’s documentation and sample result rather than transferring another manufacturer’s response or mounting values.
A standard color mark sensor is a process-detection device, not personnel-protection equipment. Registration testing does not replace machine safeguarding or safety-function validation.
Sources and method references
- SICK: Designing registration marks — scan-track artwork and contrast versus color discrimination.
- Banner: R58E Expert manual, Rev. F, June 12, 2026 — source selection and teach methods, pp. 1–2; model-specific mounting and backing, p. 3.
- OMRON: E3NX-CA application overview and ratings and specifications — packaging tasks, sensing modes and mode-dependent response.
- Beckhoff: MC_TouchProbe — an example of hardware position capture and a configurable recording window; implementation depends on the specified platform.
The worked numbers and troubleshooting scenario are illustrative, not customer test results. The scan-track drawing is conceptual and not to scale. Manufacturer examples explain selection questions; they do not establish specifications for an xsz sensor model.