Bù bán kính đầu dò CMM: nguồn gốc sai số cosin
Đầu dò ghi tâm bi chứ không phải điểm tiếp xúc; sai pháp tuyến 10° gây ~30 μm sai số cosin.

Ball center versus contact point
A touch-trigger CMM probe records the center of the stylus ball, while the measured point is the contact point on the ball surface, separated by the ball radius r, commonly 0.5 to 5 mm, along the surface normal. Software must offset r along the correct normal to recover the true surface point. The normal is inferred from the measurement strategy: plane normals for planes, bore axis directions for holes, neighboring-point fits for freeform surfaces. A wrong normal means wrong compensation.
Quantifying cosine error
When the actual approach direction deviates from the assumed normal by θ, the compensation error along the measuring axis is e=r(1-cosθ) to first order. With a 2 mm ball, 10 degrees gives about 30 μm and 20 degrees about 121 μm — far beyond machine accuracy itself. That is why slopes, cones and steep surfaces must be probed along the normal, and where five-axis articulated heads earn their keep.
Trigger pre-travel
A trigger probe needs about 5-15 μm of pre-travel from contact to signal, and lobing makes it vary by ±2-5 μm with approach direction; qualification against a reference sphere absorbs it into the effective radius per direction. Pre-travel grows with probing speed — doubling speed can add 2-3 μm — so precision work holds one speed, such as 5 mm/s. Scanning probes avoid the force snap and cut this error by an order of magnitude.
Practical rules
- Re-qualify the effective ball diameter after every stylus change or crash;
- The qualified diameter should sit within 1 μm of nominal; larger deviations mean a dirty reference sphere or a loose probe;
- On slopes prefer vector points with given normals or the sample-point method;
- On thin walls drop probing speed to 2 mm/s against dynamic error and part deflection;
- Qualify every leg of a star stylus separately — never share.



