Terms of the trade for surface roughness measurement

As the plans for more and more machined metal parts include numerical surface finish specifications, the need diminishes to visually qualify that a part has, for example, “a mirror finish with no burrs” or to have the machinist with the finest touch scratch the surface with his fingernail to make sure it’s “matte and rough.”

“When you see a print where it spells out what cutoff length, stylus tip size and filters to use, then you know that whoever put together that print really understands what he’s putting on the print and not just grandfathering it in,” said Len Carravallah, sales engineer for form products at Mitutoyo America Corp., Aurora, Ill., who works out of the company’s Plymouth, Mich., office. The company offers contact-based instruments that range from ones for performing basic measurements on easy-to-access, flat, noncontoured surfaces to automated, 5-axis CNC measuring machines.

Michigan Metrology’s Donald Cohen operates the NPFlex 3D optical profiler from Bruker. Image courtesy Michigan Metrology.

Some surfaces—such as a delicate micropart—can be negatively impacted if a stylus transverses them, and so an alternative method is needed. Bruker Corp. is one company that offers equipment for both contact and noncontact measurement of surface roughness and texture. In addition to avoiding surface damage, noncontact measurement can gather more information. “A noncontact method can enable you to get details from a surface that have to do with the surface texture direction that you may not be able to pick up with a single stylus trace,” said Matt Novak,
director of applications development for Bruker Nano Surfaces Div., Tucson, Ariz.

From Rough to Wavy

One company that uses Bruker measurement equipment is Michigan Metrology LLC. Donald Cohen founded the Livonia, Mich., company about 22 years ago and uses 3D surface microtexture measurement and analysis to help engineers and scientists solve problems related to squeaks, leaks, friction, wear, appearance and adhesion. “Those are the key issues,” he said.

Cohen uses the Bruker NPFlex white-light, vertical-scanning interferometer, which provides a high-resolution, 3D image of the surface. The surface is typically measured over a field of view of 1mm × 1mm (0.039″×0.039″). The height resolution is typically 3nm to 6nm in the vertical-scanning mode and less than 0.3nm in the phase-shifting mode.

Cohen noted a significant issue in the industry is determining the correct spatial wavelengths to measure. “In the past, everybody had stylus profilometers and they were almost hardwired to a particular cutoff length, a 0.8mm cutoff length,” he said, adding that this means any spatial wavelength, such as a sine wave on a part’s surface, that is 0.8mm (0.315″) or shorter is called roughness. Wavelengths greater than 0.8mm would be called “waviness.”

Cohen added: “The trouble is, that’s not always the case. It depends on the application. It depends on the part shape. There is nothing magical about 0.8mm as a cutoff between roughness and waviness.”

A Starrett SR series roughness tester is used to measure the finish of a machined surface. Image courtesy L.S. Starrett.

For instance, it would be a challenge to measure surface roughness on gear teeth using a 0.8mm cutoff because the part feature is often too narrow, he explained. On the other end of the spectrum, the waviness of a machined part might be measured in millimeters.

“We’re trying to get away from using words like ‘roughness’ and ‘waviness’ and just specify the spatial wavelength, usually short wavelength or long wavelength, which then defines a bandwidth,” Cohen said.

He cited a familiar example of measuring the roughness of a mountain range, which would have a roughness value that’s quite large based on the difference between the highest peak and the lowest valley. If the mountains have grass on them and someone removes the grass and measures the roughness again, the roughness measurement would be the same because the grass contributes very little to the overall roughness of the mountains.

“What I just described is what goes on in the electropolishing industry,” Cohen said. “They have a surface with some mountains on it and it might not look real shiny. Then they electropolish it, which is like removing the grass, but the big mountains are still there.”

A surface is measured with the Formtracer SV-C3200 system from Mitutoyo America. Image courtesy Mitutoyo America.

After measuring the average roughness with the typical 0.8mm cutoff, no change occurs. “That’s a huge problem that I work on throughout the year,” Cohen said. “I just polished this thing; why isn’t the roughness changing? Well, you took the grass off but didn’t get rid of the mountains.”

To solve the problem, electropolishers and other part finishers must decide whether or not those mountains are important, he said. If they are, another process is needed to remove the mountains. If not, all that matters is whether the mountains are shiny or not.

“So now, to measure the roughness, you need to measure it over shorter wavelengths,” Cohen added. “That’s filtering out the mountains. You don’t measure the mountains.”

Form is also part of the surface-texture lexicon and is an even longer wavelength, Cohen said. Form might be the shape of a part, such as a cylinder.

“When you consider the general shape, then the form is the base shape, i.e., a cylinder, and a few undulations of the surface around the total cylinder may be considered a form error, typically called ’roundness,’ ” Cohen stated. “The cylindrical shape is the form, and the deviation from an ideal cylinder would be considered the form deviation. Once you remove the ideal cylinder shape, you’re left with the deviations over some spatial wavelength that you have to define. Below that spatial wavelength is roughness and waviness.”

He added that although, historically, words like roughness, waviness and roundness have been used to describe a surface, going forward the concepts of spatial wavelength and undulations per revolution for form will be used instead.

Profile Parameters