Positioning and Tightening: How a CNC Setup Is Actually Held

• 1
  Positioning and Tightening Start Before the Cut
• 2
  Datum Choice Sets the Whole Tolerance Stack
• 3
  How Tightening Force Travels Through the Part
• 4
  Positioning and Tightening Across Multiple Operations
• 5
  When Positioning and Tightening Cannot Save the Part
• 6
  Clamping Method vs. Part Type and Risk

A CNC machine holds ±0.005 mm all day on a good setup and loses it on a bad one. The spindle is rarely the problem. The problem sits between the blank and the table: where the part was located, and how much force was used to hold it there.

Positioning is the act of restricting six degrees of freedom so the part can only sit in one place. Tightening is the force that keeps it in that place while the cutter pushes back. Treat them as one system, because a perfect locator paired with a heavy clamp will still move the part.

The 3-2-1 rule is the classic version. Three points on the primary datum kill one translation and two rotations. Two more points on a secondary face remove another rotation and translation. One point on the third face locks the last degree of freedom. Six points, six constraints, one valid position.

What engineers get wrong is mixing the order. They clamp on a rough saw-cut face and locate from a machined face, then wonder why the second op drifts 0.05 mm. The locating face and the clamping face have different jobs. Keep them separate.

On parts under 20 mm wall thickness, clamp force alone can bow the floor by 0.03 mm or more. The part springs back after unclamping and the flatness callout fails even though the machine never moved. Positioning was fine. Tightening was the error.

• 1
  Locate on machined facesRough stock carries scale, draft and saw marks that repeat poorly.
• 2
  Clamp opposite the locatorsForce should press the part into its stops, not slide it along them.
• 3
  Measure after unclampingIn-fixture readings hide elastic deflection on thin walls.

A datum is not a drawing decoration. It is the physical face the fixture will touch, and everything downstream inherits its error. When the GD&T datum and the fixture locator disagree, the CMM report and the machine report will never match.

The most common failure we see is a secondary op located on a surface that was later blended or deburred by hand. The deburring removes 0.02–0.05 mm of material unevenly. The next operation is now referenced to a face that no longer exists in the same place.

Pick datums that survive the process. A machined bore, a faced surface, or a ground pad will repeat to ±0.005 mm on a good fixture. A cast surface, a laser-cut edge, or a sheared end will repeat to ±0.1 mm at best.

For 5-axis work, the datum often has to be a feature the machine can probe. We set the rotary table center to Ø400 mm and probe a bore or a boss to establish the work offset. This keeps positioning and tightening tied to one measurable reference.

If the drawing calls for a datum that cannot be fixtured without distortion, raise it before the first cut. Redesigning a datum in CAD costs an hour. Scrapping a 10,000-part run costs a lot more.

• 1
  One datum per axisStacking three references on one face multiplies uncertainty.
• 2
  Probe what you clampIf the probe touches it, the locator should too.
• 3
  Freeze datums before productionMoving a datum after first article resets the whole stack.

Clamp force does not stop at the surface. It travels through the part in a cone roughly 45° wide, and whatever is under that cone gets compressed. If the cone reaches a thin floor, the floor bends. If it reaches a hollow rib, the rib walls splay.

A typical toggle clamp or hydraulic swing clamp applies 1–5 kN. Spread across a 20 mm × 20 mm pad, that is 2.5–12.5 MPa of local pressure. Aluminium 6061 yields around 275 MPa, so the bulk material survives easily. The geometry is what fails, not the alloy.

That is why we watch support placement more than clamp size. A jack or a self-adjusting support directly under the clamp pad turns a bending load into a compression load. The part barely moves. The same clamp with no support underneath bows the floor.

For stainless 316L and titanium TC4, springback after unclamping is larger than for aluminium. We add a roughing pass, unclamp, let the part relax, then reclamp lightly for the finishing pass. This costs one setup but recovers 0.01–0.03 mm of flatness on thin plates.

Tightening torque on the clamp bolt matters too. A 10 mm bolt torqued to 40 N·m can generate 15–20 kN of clamp force. Most operators over-torque by instinct. Use a torque wrench on the first article, then mark the position.

• 1
  Support under every clampDirect support converts bending into compression.
• 2
  Light clamp for finishing0.5–1 kN is often enough once the part is located.
• 3
  Relax between opsUnclamp, let residual stress settle, then re-clamp.

Most parts need two or three setups. Each one re-introduces positioning and tightening error, and the errors add up unless the datums are shared. The first op should create the faces that the second op will trust.

We plan the op sequence so that the second op locates on machined surfaces only. That usually means facing three sides in op one, then using those faces as the primary, secondary and tertiary datums in op two. It costs a few minutes of machine time and removes a large source of variation.

For long parts, up to 4,000 mm on our larger machines, thermal drift becomes part of the positioning problem. A 4,000 mm aluminium plate grows about 0.09 mm over a 10 °C shop swing. If the part is measured hot and assembled cold, the fit changes.

We record the shop temperature at first article and at final inspection. When the tolerance is tight, both readings need to sit within a few degrees of each other. This is not a machine limitation. It is physics.

The takeaway is that positioning and tightening are process decisions, not machine settings. A shop that treats them as an afterthought will chase tolerance in the control room. A shop that plans them will hold tolerance on the floor.

• 1
  Create datums in op oneFace three sides before any feature that needs them.
• 2
  Log shop temperatureNote it at first article and at final inspection.
• 3
  Re-probe after re-clampRe-establish the work offset whenever the part moves.

Some geometries fight any fixture. A part with a 0.5 mm floor and no accessible support point will move under cutting load regardless of how carefully it is clamped. The fix is a design change, not a better clamp.

Parts with no flat reference face are the second common case. If every surface is curved or angled, there is no place for a locator to sit. We can build a conformal nest or pour a low-melt fixture, but both add cost and lead time.

Very small parts, under 10 mm in any dimension, run into the opposite limit. The clamp pad is larger than the feature it holds, and the force concentrates on a few square millimeters. Vacuum or adhesive fixtures usually work better here.

When the tolerance is tighter than the process window, we say so early. A ±0.005 mm callout on a 0.5 mm wall in aluminium is not a fixturing problem. It is a drawing problem.

That conversation is cheaper before the first cut. Send the model and we will flag the features that will not hold before quoting, as part of the free DFM analysis.

• 1
  Thin floors need design supportAdd ribs or increase wall thickness.
• 2
  No flat face means a custom nestConformal fixturing adds cost and time.
• 3
  Flag impossible tolerances earlyBetter in DFM than in the scrap bin.