Surface roughness can be directly measured or inferred from optical measurements. Direct measurement gives you the details of the surface, but leaves you to infer the attendant optical effects. Measuring the optical scattering from a surface can tell you a lot about the surface roughness, but it doesn’t directly give you the details. We use both types of instruments, leaving nothing to chance. Here’s a little detail on our tools for surface roughness:
Our stylus profilometer is a Mitutoyo SurfTest SJ210. It has a 2 micron tip radius with 0.75 mN force and resolves 2.5 microns horizontally and 1.6nm vertically. It produces no visible scratch on acrylic, but has yet to be tried on softer materials. The resolution is great for plastic, but cannot distinguish between normally polished glass and superpolished. Of course, I wouldn’t want to drag a diamond stylus across a superpolished surface anyway.
For optical measurements, we use a 45° conoscopic scatterometer. It has sufficient resolution (0.05°) to give us BRDF measurements down to β = 0.001 and RMS surface roughness between 15 and 125 nm, with feature sizes between 1µm and 0.7 mm.