Central Machining Roundness Error: Causes and How to Adjust

• 1
  Roundness error symptom, likely cause and first fix
• 2
  What central machining roundness error actually measures
• 3
  Backlash and lost motion: the most common cause
• 4
  Servo gain, spindle error and geometry
• 5
  Tooling, workholding and thermal drift
• 6
  When roundness error matters most
• 7
  How to measure and adjust roundness error step by step

Roundness is the deviation of a real circular section from a perfect circle, measured in the same transverse plane of a rotating body. It is not the same as diameter tolerance, and it is not the same as runout. A shaft can sit dead on diameter and still fail roundness by 0.02 mm because the section is oval.

During turning or boring, the tool follows the interpolated path the control commands. Any error between commanded position and real position in the X and Y axes prints itself onto the part as a shape error. That shape is what we call central machining roundness error when it appears on a machining center rather than a lathe.

The number you read matters less than the pattern. A dial test indicator on the work piece, or a circular test with a master ball, gives amplitude and phase. Amplitude tells you how bad it is. Phase, meaning where the high and low points sit around the circle, tells you which axis or which mechanical element is contributing.

On our 127 high-precision CNC machines, including 16 simultaneous 5-axis machining centers, we treat roundness as a machine health indicator, not only a part feature. When a job that used to hold ±0.005 mm starts drifting, the machine is telling us something before the scrap report does.

Backlash, or reverse clearance, is the small dead zone when an axis reverses direction. The drive motor turns, the screw turns, but the table waits for the nut to take up the gap. In closed loop the control sees the axis lagging and commands extra movement. The circle then flattens at the quadrant points where each axis changes direction.

The signature is a two-lobe or four-lobe shape at the quadrants. You see it on interpolated circles and on any profile where X and Y reverse under cutting load. Straight-line cuts look fine, which is why operators sometimes blame the program instead of the machine.

Mechanical sources: ballscrew preload too tight or too loose, a worn nut, a coupling that has slipped, or a screw end bearing that has lost preload. Thermal growth makes it worse as the screw warms, so a machine that passes in the morning can fail after two hours of roughing.

Do not simply increase compensation and move on. If the backlash is mechanical, compensation hides it at low feed and makes it worse at high feed, because the lost motion changes with direction and load. Find the mechanical gap first, then compensate the residual.

• 1
  Check firstPush the table by hand with the servo off and a dial indicator on the slide.
• 2
  Common trapCompensating a worn nut masks the fault at low feed only.
• 3
  Thermal linkBacklash grows as the ballscrew warms. Test hot, not cold.

When one axis is tuned hotter than the other, the two axes do not arrive at the commanded point at the same time. On an arc, the faster axis overshoots while the slower one lags. The result is a four-lobe square or, at high feed, a circle that is visibly smaller than commanded. This is a tuning problem, not a mechanical one.

Servo mismatch usually appears as a feed-dependent error. Cut the same circle at 500 mm/min and at 2,000 mm/min. If roundness changes with feed, look at position loop gain, feed-forward and acceleration limits before you touch the screw.

Spindle error shows up differently. If the spindle axis is not square to the table, a bored hole comes out tapered or the circle becomes an ellipse aligned with the machine axes. Spindle bearing wear adds a once-per-revolution component that repeats at spindle speed, not at axis reversal.

Geometry faults are the slow ones. Guide rail wear, a column that has moved after a crash, or a bed that has settled over years all shift the relationship between the tool and the work. These need a squareness and parallelism check, not a parameter change.

Not every roundness error is the machine. A boring bar that is too long for its diameter deflects under cut and springs back, leaving an oval or a tapered bore. The rule of thumb is to keep the bar overhang under four times the bar diameter. If you cannot, reduce depth of cut and take a spring pass.

Workholding distorts thin-wall parts. A three-jaw chuck or a vise clamped hard pushes the bore out of round while it is being cut, and the part springs back oval when released. Measure the part free, not in the fixture. If roundness improves after unclamping, the fixture is the cause.

Thermal drift is the quiet one. A spindle that has run for an hour grows, and the tool center moves. Parts machined at 08:00 and at 14:00 can differ by more than the tolerance band. A 20–30 minute warm-up cycle before the first finishing pass removes most of this variation.

Chip control matters too. A chip caught between the insert and the bore wall adds a random bump each revolution. It looks like a machine fault but disappears after a tool change and a clean-up.

Roundness error is critical wherever a circular fit has to seal, rotate or slide. Bearing bores, hydraulic cylinder bores, spindle housings and valve seats all depend on it. A 0.02 mm oval bore can pass a diameter gauge and still leak or vibrate in service.

It matters less on parts that are located by a flat face or a pin. A bracket with a clearance hole does not care about two-lobe error. Knowing which features are functional saves you from over-adjusting the machine and burning production time.

For rotating parts, roundness error feeds directly into vibration. A shaft with a two-lobe section will show a twice-per-revolution vibration at running speed. If the application has a vibration spec, measure roundness before assembly, not after the customer complains.

In our Dongguan and Singapore plants, we inspect roundness on functional bores and journals at final inspection, with reports available on request. Parts ship in 3–5 days from a 12-hour quote, so a roundness problem caught in process is far cheaper than one caught at the customer.