What Is Backlash in a CNC Machine?
Backlash is the lost motion you feel when an axis reverses direction. It is not one fault but a stack of small gaps. This page explains where the gap sits, how to measure it, and when backlash in a cnc machine starts to matter for a 4,000 mm part or a ±0.005 mm bore.
Where backlash in a cnc machine actually sits
Backlash in a cnc machine is the small amount of travel the servo command produces before the table starts to move in the new direction. Picture a ball screw driving an X axis. The motor reverses, but the ball nut has to cross the clearance between the ball and the raceway before it pushes the table the other way. That dead travel is backlash.
The gap rarely lives in one place. It is a stack: ball-to-raceway clearance inside the nut, axial float in the bearing pair, play in the coupling between motor and screw, and a little wind-up in the screw itself over a long span. On a worn machine you can add stick-slip in the linear guides. No single part is usually out by much. Together they add up.
This is why a dial indicator on the table can show 0.02 mm of lost motion even when every component passes its own spec. The number you measure is the sum of the chain, not the worst single item.
On a rack-and-pinion axis, the gap sits between the pinion teeth and the rack teeth instead. The same reversal logic applies, but the magnitude is often larger and changes as the pinion wears.
How to measure and what number is normal
Clamp a dial indicator or a 1 μm test indicator against the table or spindle nose, parallel to the axis you want to check. Jog the axis one way by 0.5 mm, zero the indicator, then jog back past the start point. Read the indicator at the moment the table begins to move. That reading is the total lost motion for that axis and direction pair.
Repeat three times, then check the opposite direction pair. If the two numbers differ by more than about 30%, the fault is likely in one bearing or one coupling rather than uniform wear across the screw.
For reference, a healthy ball screw axis on a VMC usually shows 0.005–0.015 mm of reversal error. A larger gantry or a long-travel machine can sit at 0.02–0.03 mm and still hold its work. Lead screw axes on older manual-style machines often run 0.05 mm or more, which is why they are rarely used for tight bores.
Temperature matters. A screw that is warm after two hours of roughing will read differently than a cold one. Measure in the state you actually cut in.
Why it shows up on the part
Backlash hurts most when the tool load reverses direction. A climb-milling pass pushes the cutter one way. When the tool exits a corner and the axis reverses, the screw must take up the gap before the cutter really moves. The tool pauses, then jumps. On the part this reads as a witness mark at the corner or a step in the wall.
Boring is more sensitive than milling. A boring bar feeding out of a hole sees almost no radial load reversal, so backlash mostly affects the depth position, not the diameter. A slotting or pocketing cycle with lots of direction changes will show the error on every wall.
Surface finish follows the same pattern. The jump adds a small chatter mark, which can push Ra from a clean 0.8–1.6 μm finish up past 3.2 μm in the worst spot, even if the rest of the face is fine.
Circle interpolation is a classic case. The controller blends X and Y moves, and any reversal error in either axis turns a round hole into a slightly lobed one. On a 50 mm bore you may see a 0.02 mm lobing that no amount of cutter compensation removes.
Compensation, and where it stops working
Most controls let you enter a backlash value per axis and per direction in the parameters. The controller adds that value to the commanded move whenever the axis reverses. Done well, it removes most of the constant part of the error. Done badly, it makes things worse.
The limit is that compensation is a fixed number. Real backlash changes with load, temperature, and position along the screw. A value tuned at the home position may over-correct in the middle of travel. If you set it too high, the axis overshoots on every reversal and you get a step the other way.
Mechanical fixes hold up better over time. Preloaded ball nuts, angular contact bearing pairs with a defined preload, and a rigid coupling remove the gap at the source instead of hiding it in software. A preloaded double-nut screw can hold its setting for thousands of hours.
Compensation is still worth using. It buys accuracy between rebuilds. Just do not treat a growing compensation value as normal. When the number keeps climbing, the hardware is telling you something.
When it matters and when it does not
If your drawing allows ±0.1 mm and the part has few direction changes, most of this is noise. A 0.02 mm reversal error sits well inside the band. Run the job and check the first part.
The picture changes fast as tolerance tightens. At ±0.005 mm, the reversal error is a real share of the budget. It competes with thermal growth, tool wear, and spindle runout. You cannot spend the whole allowance on one source.
Geometry matters as much as the number. A part with long straight cuts in one direction will hold up better than a part full of pockets, slots, and small radii, even at the same tolerance. Direction changes are what expose the gap.
Material cuts in too. Aluminium at 6061 or 7075 lets you take a light finishing pass with low load, so the reversal is gentle. Inconel and 17-4PH push harder, and the screw deflects more under load, which makes the effective backlash larger than the static measurement suggests.
This is the point where a shop that measures and tracks axis health earns its keep. At GreatLight we run 127 high-precision CNC machines across 3 wholly-owned plants, with 16 simultaneous 5-axis centers, and we log reversal error as part of routine machine checks rather than waiting for a scrap event.
Drive types and their typical reversal error
Values are typical shop-floor readings, not guarantees. Measure your own machine.
| Drive type | Typical reversal error | Best fit | Watch out for |
|---|---|---|---|
| Preloaded ball screw | 0.005–0.015 mm | Tight tolerance milling and boring | Preload loss after long service |
| Standard ball screw | 0.01–0.03 mm | General milling, most prototype work | Growing number between rebuilds |
| Lead screw | 0.05 mm and up | Roughing, soft material, loose tolerance | Poor finish at corners |
| Rack and pinion | 0.02–0.08 mm | Long travel gantry axes | Gap grows as pinion wears |
| Direct drive rotary | Under 0.005 mm equivalent | 4th and 5th axis indexing | Higher cost, more tuning |
The short version
If your tolerance is looser than ±0.05 mm, track backlash but do not chase it. If you need ±0.005 mm on a part with pockets or circular interpolation, fix the hardware first and use compensation only as a top-up.
Backlash questions engineers ask
Can backlash be zero?
Not in a real machine. Any sliding or rolling contact needs some clearance to move. Preloaded ball nuts and matched bearing pairs reduce it to a few micrometres, but the number is never zero.
What you can do is hold it stable. A preloaded assembly that stays at 0.005 mm for a year is worth more than a fresh machine that drifts every month.
Does backlash affect hole diameter or position?
Mostly position and shape. A boring bar feeding straight out sees little radial load reversal, so the diameter is set by the tool, not the screw.
Round holes cut by circular interpolation are different. Reversal error in X or Y turns the circle into a slight lobe. That is a shape error, and cutter compensation will not remove it.
How often should we check axis reversal error?
At commissioning, after any crash, and on a fixed schedule. Many shops check monthly on production machines and more often on machines running tight-tolerance work.
The trend matters more than the single reading. A jump from 0.01 mm to 0.02 mm in one month points at a specific fault. A slow climb over a year is normal wear.
Will backlash show up on a surface finish report?
Sometimes. A reversal under load leaves a small mark at the corner or at the end of a cut. On a face that should read Ra 0.8–1.6 μm, that mark can push the local reading past 3.2 μm.
The rest of the surface can still be fine, which is why finish problems from backlash look local rather than uniform.
Is software compensation enough for tight work?
No. Compensation assumes a fixed number for the whole axis. Real backlash varies with load, temperature, and position along the screw.
Use it to cover the constant part. Fix the mechanical gap when you need ±0.005 mm or better on parts with frequent direction changes.
What materials make backlash worse?
Hard, high-load materials. Inconel, 17-4PH, and titanium push the cutter harder, so the screw deflects more under load. That adds to the effective reversal error.
With aluminium alloys like 6061 or 7075 you can take a light finishing pass, keep the load low, and the reversal stays gentle.
Send us the drawing and the tolerance
We quote and return a free DFM analysis within 12 hours. Tell us the tolerance band and the geometry, and we will say whether axis reversal error is a real risk for your part.
12-hour quote100% inspectionNDA on request
More machining notes
We publish setup notes, tooling trials and inspection data from the factory floor.
