Basic Knowledge of the Measurement Machine With Three Coordinates

A coordinate measuring machine records discrete points on a surface, each with an X, Y and Z value. Software then fits a line, circle, plane, cylinder or freeform surface through those points and reports size, form and position. That is the whole idea. The engineering sits in how the points are taken and how the fit is constrained.

The chain has three parts: the mechanical frame and its scales, the probe that touches or scans the part, and the software that fits geometry. Error enters at all three. A granite bridge, a temperature-stable room and linear scales set the floor. The probe sets the contact. Software decides what the numbers mean.

Keep the chain in mind when you read any spec sheet. Claimed accuracy on paper assumes a stable 20 °C room, a clean part and a probe that was qualified that shift. Break one link and the reported value drifts. Most inspection disputes we see are fixturing or thermal problems, not a broken machine.

Touch trigger probes take one point per contact. They are fast to program, cheap to replace and good enough for prismatic parts where you need position and size, not form. Cycle times stay short and the stylus is simple. For a bracket with 20 holes, this is usually the right tool.

Scanning probes drag a stylus across the surface and collect hundreds of points per second. That density is what you need to see lobing on a bore, flatness on a sealing face or profile on a cam. The trade-off is time and a more delicate stylus. Thin walls and soft aluminium can deflect under continuous contact force.

Optical and laser probes measure without touching. They suit soft parts, thin sheet, small features and surfaces where contact would mark the finish. The limit is what the sensor can see: steep walls, deep pockets and shiny surfaces cause dropouts. Many shops run a touch probe and an optical sensor on the same machine and pick per feature.

The choice is per feature, not per part. A housing might be scanned for bore roundness, touched for hole position and measured optically for a delicate rib. Mixing sensors on one program is normal on modern machines.

Spec sheets usually give two uncertainty formulas: one for a single axis and one for spatial measurement. The single-axis number is smaller and easier to quote. The spatial number is what applies when you measure a hole at an angle or a plane away from the probe tip. Use the spatial figure when you judge whether a feature passes.

Repeatability is a different animal. It describes how tightly the machine returns to the same point, not how close that point is to truth. A machine can repeat to 1 μm and still be 5 μm off in absolute terms. Repeatability is what you rely on for process control; absolute accuracy is what you rely on for a first-article report.

Uncertainty adds up. Take the machine spec, add the probe and stylus contribution, add thermal drift over the measurement time, add fixturing flex. A 1.8 + L/350 μm machine spec can become 4 μm of real uncertainty on a warm shop floor. The rule of thumb is a 4:1 to 10:1 ratio between your tolerance and your measurement uncertainty.

If the tolerance is ±0.05 mm, a CMM with 3 μm uncertainty is comfortable. If the tolerance is ±0.005 mm, the same machine is marginal and you should be looking at a temperature-controlled room, a shorter stylus and a rigid fixture. That is the judgment call the numbers are meant to support.

The part must not move. A loose part adds error that no software can remove, and it is the most common cause of a failed check. Clamp on a non-critical face, keep the clamp force low on thin walls, and support the part where the machining forces acted. A part that springs back after unclamping is a fixturing problem, not a machine problem.

Alignment defines the datum. You set the part coordinate system from features that match the drawing datums, not from whatever surface is convenient. If the drawing calls out datum A as a face and datum B as a bore, build the alignment from those two. Aligning to a random flat will rotate the whole coordinate system and make good parts look bad.

Temperature matters more than most people expect. Aluminium moves about 23 μm per metre per °C. A 300 mm aluminium part sitting in a 5 °C gradient grows roughly 35 μm before you touch it. Let the part soak in the inspection room, and measure the room, not just the machine.

Stylus choice is the last setup decision. Keep the stylus as short and as stiff as possible. Long extensions flex and add error. A ruby ball is standard for most metals; use a different tip material for soft or abrasive surfaces. Qualify the stylus at the start of every shift.

A CMM is slow compared with a gauge. For a high-volume run where you check one bore diameter thousands of times, an air gauge or a bore micrometer is faster and just as capable. Save the CMM for the first article and for periodic audits, not for every part.

Small features push against probe access. A 0.5 mm wide slot or a hole 8 mm deep and 2 mm wide may not admit a stylus at all. In those cases a vision system or an optical comparator works better. Check probe access before you write the inspection plan, not after.

Very large parts are also a poor fit for a bridge CMM. Parts beyond the machine envelope need a portable arm, a laser tracker or a shop-floor measurement system. GreatLight machines parts up to 4,000 mm, so large parts often go to a portable system rather than a fixed CMM.

Finally, a CMM does not fix a bad drawing. If a datum is missing or a tolerance is impossible to verify, the measurement will be ambiguous no matter how good the machine is. Resolve the drawing first.