Unstable Web Tension: Causes and Solutions

Unstable web tension is the fault operators describe as "the tension just will not hold": the gauge swings, rolls telescope, register drifts, or the web wrinkles at speed but runs fine at crawl. The instinct is to retune the controller, but on a real converting line most tension instability starts outside it — in roll geometry, a slipping nip, or the measurement itself.

This guide is for slitters, rewinders, presses, laminators and coaters. It assumes you know the building blocks in What Is a Web Tension Control System? and focuses on diagnosis: what to look at, in what order, and which fixes actually hold.

Rule Out a False Reading First

Before touching anything mechanical, confirm the tension is genuinely moving; a surprising share of reports are measurement problems.

1. Zero drift. Re-zero the loop with the web slack. A sensor that crept 5% overnight reports a swing the web never felt.
2. Wrap-angle change. Sensor force is F = 2 × T × sin(θ/2) plus roller weight and bearing drag, so the same true tension reads differently at full roll versus core.
3. Mechanical binding. A mount touching a guard, a snagged cable, or a stiffened pillow-block seal add a friction term that comes and goes.
4. Electrical noise. VFD cables sharing a tray with a millivolt-level sensor add a periodic component that looks like oscillation — except it stays when the line is stopped.
5. Wrong range. A sensor sized for peak tension resolves poorly at the low end; running tension under about a tenth of capacity is mostly noise.

A pillow block tension sensor or cantilever tension load cell that passes all five can be trusted. One that fails any should be corrected first.

Match the Pattern to a Source

Unstable tension is four fault families, each with a distinctive signature. Identify the family before hunting the cause.

Trace signature	Most likely source	Look first
Clean repeating cycle, fixed period	Rotating element — eccentric roll, out-of-round core, bent idler	Compute roll frequencies
Swing that grows each cycle	Control loop unstable, or actuator saturated	Output %, gain
Slow drift over 10–30 minutes	Thermal change, or roll-diameter error	Brake/clutch temp, diameter
Step at splice, roll change, accel	Inertia or taper compensation	Ramp rates, taper
Random spikes, no period	Slip, bearing damage, static	Nips, bearings

Mechanical Causes: Let the Frequency Name the Roller

The most useful floor diagnostic is period matching. Every rotating element disturbs tension once per revolution, so its disturbance frequency is f (rev/min) = V ÷ (π × D) with V in m/min and D roll diameter in metres.

Worked example: a 600 mm unwind roll at 120 m/min turns at 120 ÷ (3.1416 × 0.6) = 63.7 rpm, or 1.06 rev/s — a period of about 0.94 s. A 150 mm idler at the same speed turns at 254.6 rpm, a period of about 0.24 s. A 0.94 s cycle points at the unwind roll; no gain tuning removes it. A 0.24 s cycle points at the idlers.

Common mechanical sources, in the order found:

• Out-of-round or eccentric unwind roll. A core not concentric with the wound material changes effective radius every revolution.
• Unbalanced or bent idler rolls. Invisible at 30 m/min, they dominate at 150 m/min.
• Flat spot on a rubber-covered nip roll. A once-per-revolution tension pulse operators call "a thump every few seconds".
• Inconsistent nip pressure. A pneumatic nip whose supply sags, or whose cylinder stiction differs loading versus unloading, creates tension steps that mimic control faults.
• Bearing damage. Repetitive content at bearing fault frequencies, many times roll frequency — a period far shorter than any circumference suggests is the clue.
• Internal slip in a soft-wound roll. Layers slip inside the roll during unwinding, giving stick-slip spikes. It is a wound-roll quality problem, not a machine problem, and is routinely misread as controller gain.
• Web path misalignment. A web entering a roller off-square changes effective wrap and drag. Where misalignment is visible, treat it as guiding: see the XWDGM web guiding system.

Control and Actuator Causes

Once geometry is cleared, four issues account for most of what remains.

Gain is a mask, not a fix. Raising gain against a real periodic disturbance makes the oscillation bigger; lowering it lets drift through.

Actuator headroom. Torque demand follows T = F × R, so an unwind needs the most torque at full roll and the least at core. A magnetic powder brake or magnetic powder clutch sized only for core torque saturates at full roll and the loop runs out of authority. Read controller output at both roll extremes: above about 90% at full roll, or below about 10% at core, the sizing is wrong.

Two speed masters in one zone. Two driven points fighting produce a slow, growing swing. The rule that each zone needs one speed master and one torque source is covered here.

Diameter tracking error. Whether the controller measures roll diameter or accumulates material thickness, an error here shows as drift that worsens as the roll builds. Open Loop vs Closed Loop Tension Control compares the two methods and where each fails.

Material and Process Causes

Some instability is the material, not the machine. Stretchy films relax after a tension step, so tension downstream of a nip differs from tension a few metres later. Paper and nonwoven stiffen with humidity, so a line stable in the morning drifts after lunch.

A 20-Minute Diagnostic Sequence

Run in order; do not skip step 2.

1. Log 60 seconds at constant speed. Note amplitude and period.
2. Verify the reading. Slack the web, re-zero, and compare against a second independent measurement if possible.
3. Match the period to a roll. Compute V ÷ (π × D) for every roller in the zone and find the match.
4. Isolate the actuator. With the web stationary and the brake or clutch energised at the working setpoint, watch torque feedback. Oscillation with no web moving means the fault is in the actuator, its supply or the drive.
5. Bump test. Step the setpoint 5% and watch the response. Growing overshoot means too much gain; a slow response means too little or a saturated actuator.
6. Check headroom at both roll extremes. Record controller output at full roll and at core.

Fixes That Hold

Permanent fixes change geometry, sizing or measurement.

• Move the sensor closer to the disturbance rather than retune around a long free span.
• Fix the roll, not the gain. Replace out-of-round cores, rebalance idlers, dress flat-spotted nip covers.
• Re-size the actuator for the full roll, not the core. Where panel space is the constraint, a compact controller such as the XW micro tension controller TC200 keeps the loop without a cabinet rebuild.
• Add an isolation boundary where two processes need different tensions — a zoning decision covered in How to Choose a Web Tension Control System.
• Upgrade measurement only where it pays. Closed-loop packages around an automatic web tension controller with a through-shaft web tension sensor make sense in the zone where tension error becomes scrap, not everywhere.

FAQ

Why does tension oscillate more at high speed? Most mechanical disturbances scale with speed. Roll eccentricity produces a once-per-revolution force whose frequency rises with line speed, so past a point the actuator can no longer follow it. Oscillation that only appears above a specific speed points at a rotating-element frequency crossing the actuator bandwidth.

Why is tension stable for 20 minutes then drifts? Thermal drift. A magnetic powder brake or clutch warms to equilibrium and its torque for a given current shifts as it does. Diameter tracking error shows the same symptom but tracks roll build, not elapsed time.

Get the Root Cause Fixed, Not Masked

Unstable web tension has a few root causes and many tempting workarounds. Work the sequence above — confirm the reading, match the period to a roller, isolate the actuator, then check headroom — and you will usually name the source inside twenty minutes.

XW Machinery builds and supplies the full tension chain: automatic web tension controllers, pillow block tension sensors, cantilever tension load cells, magnetic powder brakes and magnetic powder clutches. Send us your line speed, material, roll diameters and the tension trace you logged, and we will size the actuator for your full-roll torque and tell you whether the problem is measurement, geometry or sizing.