Edge Detection Sensors for Sheet, Film and Web Position
For continuous web guiding, choose a sensor that reports lateral position, not just material present. Start with optical measurement for a clean opaque edge and evaluate ultrasonic measurement for clear film. Then match the measuring field, reference feature and output to the guide controller. The sensor measures the error; the guide mechanism corrects it.
What does an edge detection sensor actually measure?
A measuring edge sensor locates a material boundary within its sensing field. On a moving web—a continuous strip of film, paper, foil or fabric—the useful value is usually sideways position across the machine. A switching edge detector instead reports whether the boundary has crossed a threshold.
For an opaque web, a measuring optical fork or array can determine position from how much of its light field is obstructed. An ultrasonic fork uses sound attenuation. Neither should be confused with a standard fork switch that only changes state at one point.
- MeasureThe sensor reports the actual edge or printed-reference position.
- CompareThe controller compares that value with the required position.
- CorrectAn actuator moves the guide or winding station.
- Measure againThe changed web path reaches the sensor and closes the feedback loop.
Roller alignment, tension and traction still determine how the web responds to the commanded correction.
A presence signal is not a position value. PNP, NPN and relay outputs can serve web-break or limit detection. Some purpose-designed systems use three-point switching control, but that is a different controller architecture—not a drop-in substitute for proportional position feedback. E+L’s sensor range includes both measuring and three-point designs.
Should you follow an edge, the centerline or the print?
Follow the feature that determines product quality. Use one physical edge when that edge must stay aligned, both edges when center or width matters, and a printed reference when artwork position—not film position—is the requirement.
One fixed edge does not always mean a fixed center
Measure the left and right edges in the same machine coordinate system, with values representing the same cross-section or properly synchronized material location:
Width W = xR − xLxL and xR are edge positions, not unconverted readings from two separate sensor zeros.
Illustrative example: edges at 0 mm and 500 mm give a 250 mm center. If width increases to 502 mm while the left edge stays at 0 mm, the center moves to 251 mm. Single-edge guiding has met its own target, but the center has shifted by 1 mm. Center guiding would need both edge positions; it does not itself correct the material’s width.
Printed-line guiding and registration triggering are different jobs
A line-guiding sensor measures sideways movement of a line edge, line center or learned pattern. A registration sensor may only generate a pulse when a mark passes along the travel direction. A color-mark switch therefore does not automatically provide the position information needed for lateral steering.
Specify which feature must remain aligned and what happens when it disappears at a splice or unprinted gap. The Fife GUIDELINE manual distinguishes line, pattern and edge-selection modes. For discrete triggering, see the separate packaging color-mark sensor guide.
For a discrete sheet, lateral offset and angular skew are also separate quantities. Two longitudinally separated edge measurements, or a calibrated camera view, can reveal skew. Do not use the left/right width calculation as an angular measurement.
Which sensor should you try for your material?
Start with how the actual web interacts with the sensing field. A clean opaque edge is a good candidate for measuring optical sensing; clear film makes ultrasonic sensing worth testing first. Porosity, edge texture and coatings can change that choice.
| Material or constraint | First option to evaluate | What can change the choice |
|---|---|---|
| Opaque paper, card or film | A measuring optical fork, opposed array or suitable infrared edge sensor. | Holes, variable opacity, curl, dust and splice tape can alter the detected boundary. Test the complete edge, not one flat sample. |
| Clear or translucent film | An ultrasonic measuring fork, provided the material sufficiently attenuates sound. | Acoustic transmission, wrinkles, doubled layers, splice construction and available clearance. Specialized optical measurement is another option. |
| Reflective foil or metal strip | Opposed optical measurement; a qualified ultrasonic fork or optical micrometer where the application supports it. | Burrs, camber, vibration, oil and thermal movement. Opposed sensing avoids depending on a reflected return, but still needs a reliable edge shadow. |
| Nonwoven, mesh or a fuzzy edge | An optical array with suitable edge processing, or vision for the required contour. | Decide whether loose fibers, holes or a selvage belong to the controlled edge. Temporal filtering alone may hide real lateral movement. |
Treat material-based selection as a starting guide, not a guarantee. The Fife SE-44 manual, for example, requires controller-based contrast calibration for acoustically transparent material.
Clear film does not rule out optical measurement. Banner documents a specialized low-contrast EZ-ARRAY for translucent material; that capability should not be assumed for every optical fork. Its carpet application also shows why the selvage edge and tufting edge may need to be distinguished.
Use a wider field when the task needs it
A compact fork may be sufficient for one edge near a fixed setpoint. Wider arrays, measuring light grids or cameras become useful for large wander, format changes or several boundaries. Fife’s DSE-17 wideband sensor tracks multiple edges; the important selection questions are which transitions it reports, whether both required edges fit, and how missing edges are identified.
A laser micrometer can suit precision shadow measurement, but “laser” alone says nothing about total installed accuracy. Vision adds feature selection, lighting and calibration requirements. Neither is automatically preferable to a simpler measuring fork.
How much measuring range and accuracy do you need?
The usable measuring interval must contain every required edge position, including startup and format changes, with a justified allowance for uncertainty. Check where that interval sits as well as how wide it is. The fork opening is a separate mechanical dimension.
Documented model example: E+L specifies ±3 mm measuring range for the PoE FX 42, with fork-width options of 30, 60 and 124 mm. Thus a 30 mm fork option does not provide 30 mm of lateral measurement; ±3 mm spans 6 mm. These are E+L specifications, not xsz sensor ratings.
Check both end positions before adding allowances
Use one coordinate reference to establish the most negative and most positive edge positions. Include loading, wandering and width-related movement once. If a recorded envelope already includes a width change, adding the same width allowance again would double-count it.
A 13 mm span can still be misplaced
Suppose the required edge positions run from −4 to +7 mm. Add an assumed 1 mm allowance on each side: the required interval becomes −5 to +8 mm, a full span of 13 mm.
If the measuring-field midpoint can be placed at +1.5 mm, a ±6.5 mm field covers that interval geometrically. If the field must remain symmetric about the original zero, it needs at least ±8 mm, or 16 mm full span. Neither result proves usable performance at the endpoints; check the specified valid zone and sample behavior.
The allowance is an assumption for this example, not a universal margin. The lesson is to compare lower limit, upper limit and mounting reference, not just two advertised range numbers.
Do not turn resolution into an accuracy claim
Resolution describes a distinguishable change; repeatability concerns the spread of repeated readings under the stated conditions. Linearity, calibration and drift also affect measured position.
A fine sensor increment does not guarantee the required position at a downstream knife or print station. Set that process tolerance first, then evaluate the sensor and complete guide against it.
Which output will work with your guide controller?
Use the measured-position interface the controller actually accepts, with matching scaling, direction, update behavior and invalid-signal handling. A shared connector or the label “digital” does not establish compatibility.
For an analog connection, confirm the voltage/current range, input loading, supply arrangement and position endpoints. For a digital connection, confirm the protocol, process-data meaning and supported sensor/controller combination. Follow the connection drawings rather than assuming pin assignments. The 0–10 V versus 4–20 mA comparison explains the electrical trade-offs.
Under this assumed mapping, 4 mA = −10 mm, 12 mA = 0 mm and 20 mA = +10 mm. A 15.2 mA signal corresponds to +4 mm:
x = −10 + [(15.2 − 4) / 16] × 20 = +4 mm.
That result is valid only for the stated mapping. A mirrored sensor or reversed output polarity can reverse the sign. Values outside the valid range must not be silently clamped to a plausible edge position.
During the machine’s prescribed commissioning procedure, confirm that a known change in edge position produces the expected reported direction, and that the commanded correction reduces the error. Use qualified personnel and the approved setup mode; do not move or reach into a running web by hand.
Also distinguish edge outside the field, missing web, lost communication and loss of power. The controller needs a defined response to each applicable state. NO/NC output logic alone does not provide fault diagnosis or a validated machine-safety function.
Where should the sensor sit in the web path?
Mount it at the reference location specified for the guide mechanism, where the measured position relates to the required process result. Choose a stable web presentation and adequate clearance, but do not apply one mounting-distance rule to every unwinder, rewinder or steering guide.
The reference frame depends on what moves
In E+L’s ELWINDER winding-station example, the unwinding sensor mounts to the machine, while the rewinding sensor mounts to the winding station. In one case the roll feeds material toward a machine reference; in the other, the winding station follows the arriving web. Copying the bracket arrangement between those applications can change what the sensor measures.
Check the complete threading drawing: guide type, roller positions, sensor bracket reference, travel direction and point where quality is measured. A stable reading at the sensor does not establish stable position after downstream rollers. Keep fork arms clear of curl, splices and permitted flutter, and provide safe cleaning access.
Before installation or adjustment: stop and secure the machine against unintended motion and isolate energy as required by its procedure. The Fife SE-44 instructions require stopped, secured machinery for mechanical settings and warn against reaching near moving webs and rollers.
Separate material travel time from the complete control response
Illustrative timing: at 3 m/s, material takes 0.9 m ÷ 3 m/s = 0.3 s to travel from a sensor to a process point 0.9 m downstream. This can help align records for the same material location. It is not, by itself, the guide’s complete feedback delay or a tuning rule.
Likewise, a 200 Hz update at 2.5 m/s means 12.5 mm of forward web travel per update. That is not 12.5 mm of lateral measurement resolution. Sensor filtering, communication, controller execution, actuator dynamics and the web mechanics determine the useful correction response.
Why does the web still drift or oscillate?
Separate an incorrect measurement from an incorrect correction or a downstream mechanical disturbance. Compare raw position and validity, controller error, actuator command and the actual process-point result on a shared time base before changing gain or replacing hardware.
| Observed behavior | Evidence to collect | What it helps distinguish |
|---|---|---|
| Position jumps on a stationary reference | Under a secured test setup, compare raw readings and validity before/after an approved cleaning or alignment check. | Optical/acoustic instability, contamination, bracket movement or an electrical issue—not a reason to retune the guide first. |
| Jumps follow holes, fibers or a splice | Correlate the indicated boundary with the actual material feature and selected edge-processing mode. | A feature-selection problem versus true sideways motion. Validate any averaging against the process tolerance. |
| Reading is stable; the cut or print drifts | Compare the same datum at the sensor and process point; check intervening rollers, slip, tension and print-to-edge movement. | Wrong reference or downstream motion that a finer sensor resolution cannot remove. |
| Correction oscillates or reaches its travel limit | Trend error, command and actual actuator position. Review tuning, delay, available stroke, backlash and traction with the guide documentation. | Control-response and mechanical limits. Extra filtering may smooth a display while adding delay. |
| Correction moves away from the target | Leave automatic operation using the prescribed safe procedure. Verify coordinate direction, scaling and actuator direction in setup mode. | A sign/polarity error. Do not try to cure reversed correction by increasing gain. |
Clear film looks stable stopped, but loses its edge at speed
Suppose a qualified clear-film sensor produces a stable stationary value, but its validity flag drops during a controlled speed trial. First correlate each dropout with film flutter, wrinkles, splice passage and the edge leaving the measuring band. The symptom alone does not prove that the update rate is too low.
If dropouts coincide with the edge leaving the field, investigate placement or range. If the edge remains in the field but contrast changes, investigate the material/mode combination. If the raw measurement remains valid while the PLC value is corrupted, investigate the interface. A proposed replacement still needs the same trial; none of these branches is a measured success claim.
When a disturbance follows a motor, drive or cable movement, include the electrical path in the investigation. Use the sensor EMI troubleshooting guide alongside the optical or ultrasonic checks.
What should the production trial confirm?
The trial should establish both a valid position measurement and the required result at the process point, using the actual material, controller and guide settings. A clean bench signal, a high resolution figure or one fault-free pass is not enough to establish production capability.
- Define the result and test conditions.Record the controlled datum, permissible error and where it is measured. Include normal running, startup, speed changes and the actual material/format range.
- Verify the measuring interval before automatic control.In a secured setup, check positions across the required band, scaling, direction, signal validity and edge-loss behavior. Include the relevant thickness, finish, splice and edge defects.
- Check the controller and mechanics together.Qualified personnel should verify correction direction, stroke and operating modes using the machine’s commissioning instructions. Keep protective measures in place.
- Run a staged, recorded process trial.Use the approved speed/tuning sequence. Observe the sensor, controller, actuator and process-point position together through the required operating conditions and foreseeable transitions.
- Save a reproducible configuration.Record model/suffix, accessories, mounting reference, valid range, scaling, filters, controller recipe and results. State trial duration or sample coverage and unresolved conditions rather than claiming a reliability rate from limited evidence.
Do not accept the selection yet if the required edge leaves the valid band, a wrong edge can be selected without detection, invalid readings become plausible positions, or process error remains outside the agreed tolerance. Resolve the specific measurement, control or mechanical gap and repeat the affected checks.
For an application discussion with xsz sensor, share representative material samples, the required datum, the edge-position envelope, a dimensioned web-path sketch and the controller’s input requirements. Include the result required at the cutter, printer, coater or winding station. This information is more useful than asking only for “an edge sensor for film.”
This is process-sensing guidance. An ordinary edge sensor is not a personnel-protection device. Moving webs, sharp strip edges, nip points and actuators require the machine’s protective measures and appropriate safety validation.