An Expert in Decoiler Feeder & Metal Coil Processing

Home / All / Feeder Guide / How Does Backlash Show Up in a Servo Feeder?

How Does Backlash Show Up in a Servo Feeder?

Nov 20,2026

A servo feeder that loses 0.06 mm on every reversal has backlash, and no amount of gain tuning will put that length back.

The tell is not the average part. It is the first three parts after a stop, which come out short in the same direction every time.

This is the test sequence we use on a 200 SPM press when a line starts drifting and the drive reports nothing wrong.

The Pattern That Names Backlash Before You Touch a Tool

Backlash is lost motion in a mechanical connection. The motor turns, the encoder confirms the turn, and the roll does not move until the clearance is taken up.

That produces a specific signature. The error appears on a change of direction, not on a change of length, and it repeats to within a few hundredths every time.

On the press floor it looks like this: the line runs well for an hour, the operator stops for a coil change, and the first three parts after restart are 0.05 to 0.08 mm short.

Then the line settles and holds tolerance until the next stop. Nothing in the alarm log, nothing on the drive display, and the quality sheet shows a cluster of rejects at every interruption.

Compare that with a slipping roll. Slip appears under load, worsens with oil and speed, and does not care about direction or stops.

Compare it with a tuning problem. Poor gain shows as overshoot or ringing on every move, long before the parts go out of tolerance.

Four Places Lost Motion Hides

The drive train from motor to strip has four joints, and each one can contribute a few hundredths of a millimetre.

The first is the coupling between the motor shaft and the gearbox input. A taper-lock bush that was never torqued to spec will slip under acceleration and recover under deceleration, which reads as backlash.

The second is the gearbox itself. A planetary unit with 3 arcminutes of backlash produces 0.037 mm at a 42 mm feed roll radius, which is 74 percent of a 0.05 mm tolerance before anything else contributes.

The third is the roll-to-strip interface. This is not true backlash, but it behaves like it when the roll pressure is set from a table rather than from the material.

The fourth is the encoder mounting. A coupling with a cracked disc or a loose clamp gives the drive a position that is true at the motor and false at the roll.

LocationHow it showsTest that separates itWhere it costs you
Motor-to-gearbox couplingError on direction change, worse at high accelerationMark shaft and bush, jog 20 mm, check for relative movement3 to 6 hours of downtime for a part worth 80 USD
Gearbox backlashConstant offset per reversal, same at every speedReversal test at the roll, drive disabled0.037 mm from 3 arcminutes, which eats most of the budget
Roll pressureError grows with oil and with speed, not with directionRun dry strip against oiled strip at the same lengthMarks on the strip and 0.02 to 0.05 mm of slip
Encoder couplingDrive reports position held while parts varyDial indicator on the roll against commanded positionParts rejected while every alarm stays green

The Ten-Millimetre Reversal Test

The test takes fifteen minutes and needs a magnetic base, a dial indicator reading to 0.001 mm, and a machine that is stopped but powered.

Mount the indicator on the frame and put the stylus on the roll surface, tangential to rotation. Command a 10.00 mm feed, then command minus 10.00 mm, and read the indicator travel.

The difference between the commanded travel and the measured travel is the lost motion. On a healthy line it should be under 0.02 mm at the roll.

Repeat five times and watch the spread. A constant 0.05 mm is mechanical clearance; a spread from 0.01 to 0.07 is a slipping clamp or a loose coupling.

Then repeat at 100 mm and at 300 mm. Backlash is an absolute number, so it looks like 0.5 percent on a 10 mm feed and 0.02 percent on a 300 mm feed.

That ratio is the reason a line can pass a 300 mm trial and fail on a 12 mm production job.

Reading Command Against Feedback

The drive already logs the answer. Most servo drives store commanded position, actual position and following error at a few hundred microseconds, and the trace is available without extra instruments.

Command a single 10 mm move and capture the trace. A backlash signature is a flat segment at the start of the move where position does not change while the motor current rises.

A slipping roll looks different. The motor and the encoder move together and the strip does not, so the trace is clean and the part is short.

That distinction decides where you spend the next hour. Backlash is a mechanical repair; slip is a pressure, oil or roll-condition problem.

What the trace showsWhat it meansWhat it cannot prove
Flat segment at move start, current risingMechanical clearance in the drive trainWhich of the four joints, without the reversal test
Clean move, short partStrip slipping against the rollWhether the cause is pressure, oil or roll wear
Overshoot and ringingGain set too high for the inertiaThat the mechanics are sound, because gain can mask clearance
Dial indicator mounted on a servo feeder roll for a backlash reversal test

How much backlash is acceptable on a 0.05 mm feeder?

Under 0.02 mm of lost motion at the roll. At 0.037 mm, a gearbox with 3 arcminutes of clearance has already used 74 percent of the tolerance.

Does a 300 mm feed hide backlash?

Yes. The same 0.05 mm reads as 0.017 percent of a 300 mm feed and as 0.5 percent of a 10 mm feed, so short feeds expose it first.

How long does the reversal test take on a stopped line?

15 minutes for the first joint, 5 minutes for each one after. Set aside 45 minutes to cover all four locations.

Can backlash be compensated in the drive parameters?

Partly. A compensation value removes the average offset but not the variation, and it will not hold at 200 strokes per minute.

At what coil width does roll pressure mask backlash?

Above 400 mm, on 0.8 mm strip, higher pressure can hide 0.03 mm of clearance by forcing the roll to bite harder. It hides it, it does not fix it.

How often should the coupling be re-torqued?

At every 2,000 hours, and after any jam. A taper-lock bush on a 19 mm shaft takes 45 Nm, and half of that is enough to slip.

Repair Sequence, Cheapest First

Work down this list and stop when the reversal test comes back under 0.02 mm.

  • Re-torque the taper-lock bush to 45 Nm and re-run the test, cost zero
  • Replace the encoder coupling disc, 30 to 90 USD, and re-check the trace
  • Reset roll pressure from the material, not from the chart on the door
  • Re-shim the gearbox mounting and check coupling alignment with a dial gauge
  • Replace the gearbox, 900 to 1,800 USD, only when the first four are ruled out

The order matters because the gearbox is the expensive answer and the least likely one. On the last six lines we opened, three were coupling torque, two were roll pressure and one was a gearbox.

FANTY builds feeders with a specified maximum of 0.02 mm lost motion at the roll, checked on every unit before dispatch, and the figure is one of the acceptance items in the handover pack.

When the Test Gives the Wrong Answer

The reversal test measures static clearance. A line can pass it at 0.015 mm and still lose length at 180 strokes per minute, because dynamic compliance is a different number.

Thermal growth is the usual false positive. A cold gearbox can read 0.04 mm of clearance that closes to 0.01 mm after an hour of running, and the repair you ordered on Monday is unnecessary by Tuesday.

Worn rolls are the other false positive.

A polished roll surface with 0.1 mm of wear across the width will slip unevenly, and the reversal test reads that as clearance because the indicator is on a surface that no longer matches the strip.

The test also says nothing about the strip. A 0.4 mm material and a 3 mm material sit differently in the same roll pair, so a figure taken on the wrong gauge is not transferable.

Finally, the test cannot tell you how the error accumulates. A feeder with 0.02 mm of clearance per move is fine on a single pitch and 0.4 mm out after 20 progressive strokes, if the die does not re-register.

The error budget that allocates tolerance between clearance, slip, back tension and pilot release is set out in our coil line knowledge base.

Roll diameters, pressure ranges and gearbox ratings for each servo feeder frame are in the feeder data library, the sheet to check before ordering a replacement unit.

Where to go next

Backlash is one line in a longer budget. The other lines, from decoiler tension to pilot release, decide whether the total fits inside your tolerance.

The accuracy pillar allocates that budget and shows which contribution is cheapest to reduce.

Read the error budget

Follow us to Get Latest Updates

subscription