Roughness average, commonly written as Ra, is the most widely used number for describing how smooth or rough a surface is. It represents the arithmetic mean of all the tiny peaks and valleys along a surface profile, measured from a center line over a set distance. If you picture a perfectly flat line drawn through a surface’s texture, Ra tells you the average distance the actual surface deviates from that line. It’s reported in micrometers (µm) in metric units or microinches (µin) in imperial units, where 1 µm equals roughly 40 µin.
How Ra Is Calculated
Ra is defined by a straightforward concept: trace a line across a surface, record every peak and valley relative to the mean line, take the absolute value of each height measurement (so valleys count as positive numbers just like peaks), and then average them all. The formal version of this is an integral of the absolute height values over the sampling length. In discrete terms, you sum up the absolute height at each measured point and divide by the total number of points.
What makes Ra intuitive is that it collapses all the complexity of a surface into a single number. A machined steel part with an Ra of 0.8 µm is smoother than one with an Ra of 3.2 µm. That simplicity is both its greatest strength and its biggest limitation.
What Ra Doesn’t Tell You
Ra is not sensitive to individual peaks or valleys. Two surfaces can share the same Ra value yet look and perform very differently. One might have a few deep scratches separated by smooth stretches, while the other has a uniform, fine texture. Ra averages these differences away. Research published in the ASME Journal of Turbomachinery demonstrated that relying on Ra alone to predict real-world behavior, like how much a rough surface increases drag, is oversimplified. The roughness slope, meaning how steeply the peaks rise and how far apart they are, matters just as much as the average height, and Ra captures neither.
This is why engineers often pair Ra with other parameters. Rz measures the average distance from the highest peak to the deepest valley across five sampling sections, making it much more responsive to scratches, damage, and contamination. Rq (also called RMS roughness) squares each height value before averaging, which gives extra weight to outliers. For critical applications like sealing surfaces or bearing journals, looking at Ra alone can be misleading.
Ra vs. Rz vs. Sa
Ra is a two-dimensional, line-based measurement. You drag a stylus or scan a laser along one path and get a single profile. Sa is the three-dimensional equivalent: instead of measuring along a line, it captures an entire area and calculates the arithmetic mean deviation across that surface. Sa is considered more statistically reliable because it isn’t dependent on where exactly you place your measurement line. Studies have found that area-based measurements (Sa) tend to produce higher values than their line-based counterparts (Ra) on the same surface.
Rz, by contrast, focuses on extremes. Because it captures peak-to-valley distances in each sampling section, it reacts strongly to a single deep gouge or tall spike that Ra would barely register. In practice, a quality inspector checking a ground surface might specify both Ra (for overall texture) and Rz (to catch defects).
How Ra Is Measured
The most established method uses a contact stylus profilometer. A small conical diamond tip drags across the surface, physically following every contour, and records height data along a 2D line. The equipment is affordable, the physics behind it are well understood, and the process is straightforward. The drawback is that the tip can scratch soft materials, it only captures a single line at a time, and repositioning it in exactly the same spot for a repeat measurement is nearly impossible.
Optical methods are gaining ground. Confocal laser scanning microscopy offers the highest resolution of the common alternatives and captures full 3D surface maps, but measurements are slow. Fringe projection is the fastest optical technique and covers large areas with minimal post-processing, though its resolution is considerably lower. All optical methods share a limitation with stylus instruments: none can detect re-entrant features (undercuts or overhangs where the surface curves back under itself). X-ray computed tomography can, but it’s far more specialized and expensive.
Importantly, different measurement devices do not always agree. Studies comparing stylus profilometers, scanning electron microscopes, and confocal microscopes found statistically significant differences in the Ra and Rz values they reported on the same surfaces. This means the measurement method should be specified alongside the Ra number for results to be meaningful.
Why Cutoff Length Matters
Before calculating Ra, the raw surface profile is filtered to separate roughness (the fine-scale texture you care about) from waviness (broader undulations caused by things like machine vibration). The cutoff wavelength, called λc or Lc, determines where that boundary falls. Everything shorter than the cutoff is treated as roughness; everything longer is filtered out.
The choice of cutoff dramatically affects the result. A surface with a “true” Ra of 20 µm, measured with a cutoff of 0.08 mm, would return an Ra of virtually zero because the filter removes all the relevant texture. At a cutoff of 0.25 mm, the same surface might read around 10 µm. Only at 0.8 mm or 2.5 mm would you get the correct 20 µm value. Set the cutoff too large, and waviness gets included, inflating the number further.
A practical rule of thumb: set the cutoff to about five times the average spacing between peaks. ISO 4288 (now largely folded into the ISO 21920 series) provides lookup tables based on expected Ra ranges and machining processes. If someone hands you an Ra value without mentioning the cutoff, the number is incomplete.
How Ra Affects Part Performance
Surface roughness directly influences friction, wear, and sealing ability. For parts that slide against each other, a rougher surface means less actual contact area, because the load concentrates on the tips of the peaks. This sounds like it would reduce friction, but in practice, those concentrated contact points wear down faster, generate debris, and accelerate damage. Research on ground surfaces found that higher initial Ra values lead to a longer break-in period before friction stabilizes, and that very rough surfaces lose material more readily during reciprocating motion.
Higher friction coefficients also translate directly to higher energy costs in mechanical systems like hydraulic pumps, where contact between vanes and housing grooves is a primary failure point. For sealing applications, a surface that’s too rough allows fluid to leak through the valleys, while a surface that’s too smooth may not retain enough lubricant. Most seal manufacturers specify an Ra window, often between 0.2 and 0.8 µm, rather than simply asking for “as smooth as possible.”
Governing Standards
The international framework for surface texture measurement has recently been consolidated. The older ISO 4287:1997, which originally defined Ra and related profile parameters, has been withdrawn and replaced by ISO 21920-2:2021. In the United States, ASME B46.1 serves a parallel role. Both standards define how to calculate Ra, which filters to apply, and how to report results. If you’re specifying Ra on a drawing or in a contract, referencing the current standard ensures everyone measures the same way.
The shift toward ISO 21920 also reflects a broader move toward areal (3D) parameters like Sa alongside traditional profile (2D) parameters like Ra. Many industries are adopting both, using Ra for backward compatibility with existing specifications and Sa for more comprehensive surface characterization.