Why Does a Ball Screw Develop Backlash? 7 Common Causes

Author:    Date:2026-10-29 

An axis that used to repeat tightly now repeats loosely, and the first conclusion drawn is that the screw is finished. Sometimes that is right. Just as often the play that appeared is not in the nut, and the nut that gets ordered is not the part that fixes it.

Growing axial play does not arrive as a failure. It accumulates — a little from wear, a little from a support that no longer holds the shaft where it should, a little from a nut that was never set to the value on the drawing — and it becomes visible only when the process window closes around it.

Treating it as a signal to buy a nut is the expensive reading of the symptom. Seven sources are worth checking, in this order, and several of them are not in the nut at all.

What this article covers. It sets out the seven sources one by one, in the order worth checking them, and gives one way to tell each from the others. It does not explain what preload is or how preload and clearance relate — that has its own article on this site — and it does not assume a new nut is the answer: in the support end and in a slipping coupling it changes nothing, and in an assembly fault the nut may not need replacing at all.

Nothing below is the specification of any part: the seven sources are described without numbers, and the only figure that governs your axis is the one you measure on it.


1. The short answer

Axial play is the free movement between screw and nut when the direction reverses: the screw turns, and the table waits for the play to be taken up before it follows. Measure it and you have a number. Where the number came from is a separate question, and the seven sources fall into three groups — wear in the rolling contact, preload that was never right or has since relaxed, and things outside the nut that let the shaft move instead.

The play can sit outside the screw and nut — in the support end, or in a coupling that slips — and either is indistinguishable from a worn nut at the tool. That is why the measurement comes before the order: placed on the symptom alone, the order is wrong often enough to matter.


2. Wear in the rolling contact

Balls and raceways are loaded through a small contact patch at high local stress, and every metre travelled takes something off them. Over a long duty cycle the material lost becomes clearance, and the preload that was holding the assembly tight has less to push against. This is the source most people mean when they say a screw is worn out, and on a machine that runs continuously it is usually present to some degree.

What decides the rate is distance travelled and load, not time: an axis that moves constantly wears far faster than one that moves twice a shift, whatever the calendar says.

Wear is also concentrated in one part. Contact is intermittent everywhere, and what differs is how often each point is passed: every point of the short nut raceway sees balls through the whole travel, while a point on the screw sees them only while the nut is over it. On inspection, the nut raceway comes back worn while the screw still measures within tolerance, provided the axis uses most of its stroke and has not been damaged by debris or a crash.

How to tell it apart from the others. The play grows slowly and evenly, and it tracks metres travelled rather than any event. A drag torque reading taken the same way each month falls as the preload relaxes. Nothing else on the machine changed.


3. A preload class that never matched the duty

The second source is not wear at all: the axis was loose from the day it was assembled. The preload class on the drawing was chosen for a duty lighter than the axis actually runs, or for a load case nobody wrote down, and the assembly has been working at the edge of its window ever since.

Nothing has degraded and nothing needs replacing. What is wrong is a specification decision made upstream, and the fix is a different class on the next order rather than a new part now. Preload and clearance, and how the classes are set out, are covered in the preload and backlash discussion.

How to tell it apart from the others. The play was there at acceptance. A reading taken during commissioning already sits outside the window the process needs, and — the part that separates it from wear — it does not grow.


4. Overload and shock loading

A single load event large enough to drive the contact stress past the elastic limit leaves a permanent dent in the raceway: brinelling. A crash, a jam, a fixture that hit a stop, or a lift that dropped onto its end can do it once. After that the balls run over a damaged patch every time they pass it.

This one is a discontinuity, not a trend, and it is judged against a different rating. Running load is what the dynamic rating Ca describes; a dent is what the static rating C0a describes — defined in ISO 3408-5 — which is why an axis can run inside its dynamic rating for years and still be marked once, at standstill, by a crash or by a load set down onto it.

That limit is the static axial load rating C0a: the load at which the contact takes a permanent set instead of an elastic one. It applies to an impact load whether the axis is stopped or moving.

How to tell it apart from the others. The play appeared after an identifiable event, and it is localised: the axis runs rough, or noisy, at one place in the stroke rather than throughout. Look for a corresponding mark on the screw at that position.


5. Lubrication: starvation and contamination

Rolling contact needs a film. Without it, or with dirt in it, the surfaces that should be separated by lubricant are separated by their own wear debris instead, and abrasive wear is far faster than normal running wear.

How to tell it apart from the others. Torque and temperature rise before the play does, and the lubricant itself is the evidence: discoloured, gritty, or dry where it should be wet. The rate is faster than distance travelled alone would explain, and the seals are usually visibly past their life.


6. Misalignment and side load

A ball screw is built to carry axial load. Parallelism between the screw and the guides, and squareness of the mounting faces, decide whether that is what it actually gets. When it does not, one side of the nut carries more than the other, the wear is uneven around the screw, and the assembly loses its geometry in a way no amount of replacement parts will fix.

The causes are ordinary ones: a guide that is not parallel to the screw, a mounting face that is not square, or a load carried off to one side of the carriage. All three put a moment on the nut that its circuits were never sharing.

How to tell it apart from the others. The damage is not symmetric: turn the shaft by hand and the resistance varies with angular position. It commonly appears after the guides or the coupling have been disturbed, and the parallelism check is cheap to run before anything is ordered.


7. The support end — where the play is not in the nut

The fixed end holds the shaft axially: an angular contact bearing pair, a housing, and a locknut. If the bearings were not preloaded, if the housing has worked loose, or if the locknut has backed off, the shaft moves inside its own supports and the axis shows exactly the same symptom as a worn nut.

This is the one most often misdiagnosed, because the symptom is identical and the nut is easier to blame. It is also the one that comes back after a repair: the play disappears when the shaft is re-clamped and returns when the housing works loose again.

How to tell it apart from the others. Mount an indicator on the machine frame with its tip on the shaft end or the locknut, and push and pull the shaft with the drive isolated, the carriage supported, and the nut flange unbolted from the carriage, so the test load cannot run back through the nut: what it reads is the support end's share. Then bolt the flange back to the carriage, put the base on the screw and the tip on the nut flange, and push and pull along the axis: that reading is the nut's.


8. Assembly: the nut was never set to the value

The last source is a fitting fault. A nut assembled without being brought to its specified setting, a two-body preload whose spacer was never locked, a coupling that slips on the shaft instead of gripping it — all of these put play into an otherwise sound axis, and all of them are found at the bench rather than in a catalogue.

How to tell it apart from the others. The play was there immediately after assembly or reassembly, and it is often inconsistent: it changes when the coupling is re-tightened, or it appears between the shaft and the table rather than between the screw and the nut.


9. Working out which one it is

Take them in this order, because the cheap checks come first and the expensive conclusion comes last.

  1. Measure the play and write down how you measured it. A dial indicator on the carriage against a reversed command gives a usable figure; the method matters more than the instrument, because the trend is what you will be reading. Record the direction you approached from and where in the stroke you took it — play that varies along the stroke is a different finding from play that is the same everywhere.
  1. Separate the support end from the nut. With the indicator on the machine frame and its tip on the shaft end, push and pull the shaft with the drive isolated, the carriage supported, and the nut flange unbolted from the carriage: movement there is play in the support end, and no nut will fix it.
  1. Check lubrication and alignment. Both are cheap to inspect, and both are causes rather than symptoms: leaving them wrong puts the replacement on the same path as the original.
  1. Only then look at the nut. Whether the play is in the rolling contact or in a preload that was never set decides what you order, and single nut vs double nut: what's the difference is where the configuration question is answered — including the cases where the answer is not the adjustable one.

10. How much play is acceptable

There is no figure this article can give you. The window is set by what the operation tolerates: an axis that always approaches from the same direction can carry a great deal of play without ever showing it, while one that reverses under load shows all of it at the tool.

What matters more than the reading is the trend. One reading is a snapshot; two taken a known interval apart are a trend. Take the same measurement the same way at a fixed interval, and it tells you whether the axis is stable, wearing normally, or accelerating — and the third of those is the one worth acting on before it becomes a stoppage.


11. Frequently asked questions

Is growing axial play always wear? No. Wear is the most common single source and the one everyone expects, but a loose support end, a coupling that slips and a preload class that was never right all produce the same symptom, and none of them is wear.

Can I measure it without special equipment? A dial indicator on the carriage against a reversed command gives a usable figure. Consistency of method matters more than the precision of the instrument, because you are reading a trend over months, not certifying a part.

Will a new nut fix it? That depends on which of the seven it is. Where the play is in the support end or in a coupling, a new nut changes nothing; where it is misalignment, a new nut will wear the same way as the old one unless the alignment is corrected first.

Does the play keep growing at the same rate? Not necessarily. Normal running wear tracks distance travelled and grows slowly and evenly. Contamination and misalignment accelerate it, so a rate that is climbing is evidence about the cause, not just about the amount.

Should the screw be replaced along with the nut? Usually not — inspect the screw before ordering one. On inspection, the balls and the nut raceway are what wear, and the screw still measures within tolerance, provided the axis uses most of its stroke. Where the screw is marked, it will be at one place in the stroke, and the mark is part of the diagnosis rather than a separate fault.


12. Data sources

  • ISO 3408 — ball screws; part 5 defines the static axial load rating C0a referred to in section 4, and the vocabulary used here matches the rest of this site.
  • Wear of rolling contacts, brinelling under a single overload, and abrasive wear from contamination or lubricant starvation — general machine-design and tribology treatment; no numerical rate is quoted in this article.
  • The seven sources and the discriminators in sections 2 to 8 — compiled from the failure modes described in maker catalogues and general ball screw maintenance references; makers group them differently, and the order here follows the cost of checking rather than the frequency of occurrence.
  • Acceptable play in section 10 — no figure is quoted; the window is set by the process, not by the screw.
  • Measurement methods in sections 9 and 11 — general workshop practice; any specific procedure should follow the machine builder's service documentation.

Nothing here is a specification of any part, and no Pinsi value appears anywhere in it. It is general technical guidance and does not replace a diagnosis done on your own axis.

Last updated: 2026-10-29


13. An axis that is losing position?

Send the stroke, the duty cycle, the load case, and the play you have measured and how you measured it. Those four tell us which of the seven you are looking at before anyone talks about parts.

top