Look at the edge of a crisp packet or a coffee sachet and you’ll find a small solid rectangle printed in the seal area, usually black, usually somewhere the design doesn’t reach. It’s called a registration mark, and it exists because printed film and the machine cutting it do not naturally stay in step.
The sensor that reads it is a specific product category, and choosing one badly is a well-known way to lose a shift.
Why printing drifts
Packaging film is printed on one machine and filled, sealed and cut on another. The print repeat is fixed at the printing stage. The film then goes onto a reel, through storage, onto the packaging machine, and gets pulled through a web path under tension.
Plastic film stretches. Not much, but consistently, and the amount varies with tension, temperature and how much is left on the reel. Over a few hundred packs the accumulated error is enough to put the cut through the middle of the artwork.
So the machine doesn’t count. It looks for the mark and corrects the film position on every repeat, which turns an accumulating error into a bounded one.
Why a normal photoelectric sensor won’t do it
A standard diffuse sensor answers “is something there.” A colour mark sensor answers “has the surface in front of me just changed contrast,” which is a different question, and three things follow from it.
It has a very small, sharply defined spot. Often around 1 × 4 mm. A large spot averages the mark and the background together, and the transition softens until the trigger point wanders.
It’s fast. Response times in the tens of microseconds. On film at 6 m/s, a 4 mm mark is present for under 700 microseconds — a general-purpose sensor with a 1 ms response cannot see it at all.
It works on contrast, with a teach function. You show it background, show it mark, and it sets a threshold between them. Absolute reflectance doesn’t matter; the difference does.
The part that gets chosen wrong: light source colour
This is the decision that separates a reliable line from a temperamental one, and it’s counter-intuitive.
A coloured surface looks coloured because it absorbs some wavelengths and reflects others. A red mark reflects red and absorbs green. So under a red light source, a red mark is bright — and if it sits on white film, which is also bright, the sensor sees very little difference.
Under a green light source, that red mark absorbs almost everything and appears dark, while white film still reflects strongly. Large contrast, clean switching.
The rule that falls out of this: choose the light colour that the mark absorbs, not the one that matches it.
Red LED for blue and green marks. Green LED for red and magenta marks. Blue for yellow, which is a hard case in general. White light sources with multi-channel receivers handle mixed jobs at some cost in ultimate contrast, and they’re the pragmatic choice when one machine runs many designs.
Manufacturers publish contrast tables showing expected performance for each source colour against common mark and background combinations. The colour mark range from GTRIC is organised this way, and reading that table before ordering is the entire difference between a sensor that teaches in ten seconds and one that needs re-teaching every reel.
Practical points
Keep the sensing distance constant. These sensors have a small depth of field and are specified at a particular distance. Film flutter changes the reflected level, so the web needs support — a roller or a plate under the sensing point.
Mind glossy film. A specular surface reflects the spot away rather than back. Mounting the sensor at a slight angle, typically around 10 to 15 degrees off perpendicular, avoids the direct reflection.
Re-teach on material change. New film, new supplier, new ink batch — the contrast has changed even if the design hasn’t.
Check the mark is actually printed to spec. Some drift faults come from the print, not the sensor: a mark that has faded across a reel or that varies in size will produce exactly the symptoms people blame the sensor for.
