Surface texture is evaluated from measured profile or areal data, but the raw measurement contains variation at different spatial scales.

A contact stylus profilometer moves a small tip across the surface and records vertical displacement. Optical methods can obtain surface height without physical contact using several different optical principles.

The measured trace must be separated into components associated with form, waviness and roughness. A cutoff or spatial filter determines which wavelength range contributes to the reported roughness.

This matters because $R_a$ is not meaningful without an evaluation procedure. The same raw surface can produce different reported values if the sampling length or filtering is changed.

Suppose a profile contains a slow 20 $\mu$m bow across several millimetres plus 1 $\mu$m tool marks every 0.2 mm. A roughness analysis should remove the slow form/waviness component rather than treating the entire bow as microscopic roughness.

Measurement direction also matters on directional surfaces. Traversing parallel to machining lay can produce a very different profile from traversing across the tool marks.

Surface metrology therefore requires specifying:

• measurement method and tip/optical resolution;
• sampling and evaluation length;
• filtering/cutoff;
• measurement direction;
• reported texture parameter.

A roughness number is the output of a defined measurement process, not a property obtained independently of scale and filtering.