You load a roll, hit print, and the first meter looks fine. By the third meter the right edge has walked inward, and when the banner comes off the take-up the two edges differ by a few millimeters across the length. Nobody notices until the hemming guy complains or the print does not fit the frame.
Here is the short answer before you pick up a wrench. Crooked output on a roll fed printer has three physical sources: the media is being pulled at an angle, the media is being gripped unevenly, or the media is not being held flat. A worn feed motor and a loose pulley are real causes, but they sit at the end of the list, and they are the only ones that cost actual money.
Start with one measurement. Print one meter of a real job at production speed, stop the carriage, and mark the media edge against a fixed reference on the platen. Note both sides. Everything below depends on whether those numbers are wrong from the first pass or get worse as the roll runs.
Split the Problem Into Two Shapes Before You Touch Anything
There are two shapes of skew, and they point at different hardware. Born crooked means the very first pass is already off. Drifting crooked means the job starts true and walks sideways as the roll feeds. Confusing the two is why people replace a feed motor to fix a loading mistake.
Run this protocol. Load the roll the way you normally do. Mark the media edge with a fine pen where it crosses a fixed reference on the platen, for example the left edge of the vacuum zone. Print one meter at the speed you actually sell. Stop, do not cut, and mark again at the same reference. Print three more meters and make a third mark.
Three marks give you a shape. Equal offset at one meter and at four meters means the media went in crooked and stayed crooked, so the fault is upstream of the print zone. Offset that grows steadily means something is slipping or pulling the media sideways while it runs. One sudden jump between two marks, with clean runs on either side, means something physical caught the media: a wrinkle passed through, an edge snagged, or a roll splice went by.
Photograph the three marks with a tape measure in the frame. That photo does two jobs. It tells you which half of this article to read, and it is the first thing a parts seller or a technician should ask you for instead of guessing.
Two more observations belong in the same note. Watch whether the drift is always on the same side, and watch whether it appears on every media or only on one roll. Same side on every roll points at the machine. One roll only, and the roll itself is now the suspect.
Measure the second edge with a steel rule against the platen, not by eye against the printed image border. Images get cropped and scaled, so the border can be wrong while the feed is perfect. Keep a strip of tape on the platen as your reference point and use the same one every time, otherwise the numbers you compare come from two different rulers.
One distinction worth making early. If the printed image is the right length but sitting at an angle on the media, that is skew. If the image is square to the media but longer or shorter than the file, that is a feed length error and it belongs to calibration rather than to the parts in this article. If both edges converge, the image is narrower at one end, you may have both problems at once, and the skew is the one to fix first because it corrupts every calibration you run afterward.
Keep the panels short while you are diagnosing. Panel length and misalignment track together, so long panels make a small error look like a big one and send you hunting in the wrong place. Short panels are a diagnostic tool, not a fix.
Free Fixes First: Seating, Rear Slack, and Take-Up Height
Roll seating is boring and it is the single most common cause. Push the roll until the core actually touches the flange on both ends, then check that the flanges sit square to the shaft. A roll hanging a few millimeters off one flange feeds at a slight angle for its whole length, and no calibration will correct that.
Check the core before it goes on. Stand the roll on the floor and roll it a short distance. A crushed or oval core hops as it turns, which changes the effective feed rate once per revolution. Look at the wound edge too. A telescoped roll, one where the layers have slid sideways, cannot track straight no matter how carefully you load it because the media is being unwound from a spiral.
The rear of the machine matters more than most operators think. A heavy roll pulls back against the feed motor, and the strain shows up as a slight skew that grows with distance. Leave slack behind the roll, and turn on prefeed if the machine has it, so the motor is not dragging the full weight of the roll through the feed path every pass.
Now the test worth doing before any other. Run three meters with the take-up disconnected and let the media hang or coil on the floor. If the skew disappears, your machine is fine and the take-up is pulling one edge harder than the other. If the skew is unchanged, tension is not your problem and you can cross the whole take-up assembly off the list.
Take-up height is easy to verify with a strip of media. Loop a long strip around the take-up core, pull it up through the pinch rollers, and lay it flat on the platen. Mark the strip at the same platen reference on the left and on the right. If the two marks do not line up, the take-up is not parallel to the platen, and most take-up arms have an adjustment screw for exactly this.
Heat belongs here as well. On banner stock, platen heat softens the media and lets it buckle between the pinch rollers and the take-up. Try the same job with the print heater down and see whether the buckling, and the drift that comes with it, goes away.
Storage and acclimation are part of the feed path whether or not the machine is at fault. A roll that sat in a cold van overnight and goes straight onto a warm platen curls as it warms, and curl on one edge is a sideways force on the whole web. Give a cold roll an hour in the room, standing upright, before it goes on the shaft. Rolls stored leaning against a wall take a set that shows up as a drift in the first two meters and then settles, which is why the first panel off a new roll is the one to distrust.
Check the wound edge against the flange at both ends after loading, not just at the near side. It is easy to seat one end and leave the other floating a few millimeters out, and the angle that creates is invisible until the third meter.
Grip and Hold: Pinch Rollers, Grit Roller, Vacuum
If the roll is seated, the rear has slack, and the take-up is off the table, you are now looking at how the machine holds the media. Two systems do that job: the pinch rollers pressing down onto the grit roller, and the vacuum pulling the media onto the platen.
Pinch roller pressure is the most common mechanical cause of one sided drift. Each wheel has to press with the same force as its neighbors. If one wheel is weak or glazed, the media on that side advances slightly less per step, and the error accumulates into a curve. Clean the wheel faces and the grit roller bar underneath before you conclude anything. Cut debris packed against the grit roller is enough to make the media meander, and a stiff brush is the right tool for that groove.
Look at the wheel surface under a light. Flat spots, a polished glaze, or a hard edge where the media rides tell you the roller is done. Wipe with alcohol if the surface is contaminated, but do not make a habit of it, because repeated alcohol cleaning dries the rubber and makes the glazing worse. If one wheel is damaged, plan on the set.
With the machine powered down, lift and release each pinch roller by hand and compare the spring feel. Then pull a strip of media through one roller pair at a time and feel the drag. One pair noticeably lighter than the others is your drift source.
Vacuum is the part people set once and forget. Too little suction on stiff banner and the edges lift, which both invites a head strike and lets the media wander. Too much suction on thin film and the media cockles into waves, which changes the path length on one side and produces the same drift. Set vacuum per media instead of leaving it on auto, and note the setting on the job sheet.
Clean the platen holes with a vacuum and a soft brush. Dust, dried ink mist, and media lint pack the holes from below over months, so the suction that reaches the media becomes uneven across the width. A local dead zone lifts the media exactly where the carriage passes, and the first symptom is usually a strike before it is a skew.
After any head strike, inspect the platen surface before restarting. A chip of dried ink or a sliver of media sitting proud of the platen holds the media up and recreates the problem on the next pass.
Watch media width against the vacuum zone. A narrow roll sitting on a wide platen leaves the outer holes open, so suction bleeds off and the media that is there gets held less firmly than the setting suggests. Tape the unused zone or pick a vacuum level for the actual width, and the drift on narrow jobs often disappears without touching a roller.
Test at production speed, not in the highest quality mode. Low pass counts and high carriage speed are where grip and suction failures show up first, so a test run at half speed can convince you a problem is fixed when it is only hidden.
The Order to Work In, and When It Is Really the Drive
Work from free to expensive. This order exists because each step eliminates a whole family of causes, and because the expensive parts are the least likely to be at fault.
- Reseat the roll against both flanges and reject crushed or telescoped rolls.
- Leave slack behind the roll and enable prefeed.
- Run the take-up off test to separate tension from the machine.
- Level the take-up against the platen using the strip and two marks.
- Clean pinch rollers and the grit roller, then compare pressure and drag pair by pair.
- Set vacuum for the media and clear the platen holes.
- Run feed calibration for that specific media and print mode.
- Only now inspect the belt, pulleys, and feed motor.
Understand what calibration can do. Feed adjustment compensates for normal variation in advance length. It cannot fix a slipping pulley, a loose belt, a contaminated roller, or an unstable motor connection. Using calibration to mask a mechanical slip is a trap: the numbers get pushed further and further, the next roll behaves differently, and eventually the correction runs out of range.
Three signs say the drive really is at fault. The skew repeats on several different rolls after correct loading. It started suddenly and now affects every media. It comes with noise, roller marks, sensor errors, or periodic jams. A steady ramp on every media with no noise is more often grip than drive.
Write the symptom timeline down before you call anyone. Machine model, media type and width, roll or sheet, the exact place where the drift starts, and any panel message. A technician with that note goes straight to the right subsystem, and a parts seller with that note stops shipping you the wrong pulley.
Prices are not the issue here, downtime is. Count it as hours of output you lost times your hourly rate, plus the media you scrapped, plus rush freight if a job is waiting, plus a service visit if the first two attempts fail. Two reprinted banners usually cost more than the roller set you were hesitating over.
One buying caution. Do not order a pulley by tooth count alone. Match pitch, bore, width, bearing, set screw position, belt profile, and shaft fit against the part you removed, and photograph both sides of the original. And treat a crooked roll as evidence of a feed path problem, not as evidence that the feed motor failed.
Change one thing at a time and reprint the same four meter test after each change. Two adjustments at once and three meters of banner is a wasted experiment, because you will not know which one did the work and you will still be guessing next month. Write what you changed and what the new marks read on the same page as the old ones.
Before you order anything, repeat the three mark measurement and the take-up off test with the machine clean. Half the crooked print tickets I have seen end there, with nothing but a wiped grit roller and a reseated roll.