Surface roughness Ra versus Rz: what each parameter measures and when to specify it
Ra averages the profile; Rz averages the tallest peaks and valleys. Same surface, two different numbers. This page shows how far apart they drift by process, how a stylus and cutoff change the reading, and which callout to put on the drawing.
Surface roughness Ra versus Rz at a glance
Values are typical ranges, not acceptance limits.
| Item | Ra | Rz |
|---|---|---|
| What it measures | Mean of absolute profile heights | Mean of the five highest peaks and valleys |
| Sensitivity to one scratch | Low, one peak barely moves it | High, one deep valley dominates |
| Typical Rz/Ra ratio, ground | About 4 to 6 | Reference value |
| Typical Rz/Ra ratio, turned | About 5 to 10 | Reference value |
| Best use | General finish, appearance, bearing area | Seals, gasket lands, fatigue fillets |
| Weak point | Hides peak height and valley depth | Noisy over large areas, single trace |
| ISO 4288 cutoff at Ra 1.6 μm | 0.8 mm | 0.8 mm, same discipline |
| Stylus that suits it | 2 μm tip is enough | 5 μm tip traces valleys better |
What Ra and Rz actually measure
Ra is the arithmetical mean deviation of the roughness profile. The instrument filters the trace, removes form and waviness, flips every height above the mean line to positive, and averages them over the evaluation length. Every point on the trace carries the same weight. A single 6 μm peak on an otherwise smooth surface moves Ra by almost nothing.
Rz works differently. Under ISO 4287 it takes the five highest peaks and the five deepest valleys inside the sampling length and averages them. The bulk of the surface is ignored. That is why Rz reacts to one bad scratch and Ra does not.
The two parameters answer different questions. Ra asks how much material sits away from the mean line on average. Rz asks how tall the worst features are in the short window the stylus crossed. Neither one describes shape, slope, or how much bearing area the surface offers.
Both numbers depend on the cutoff and evaluation length, so a bare value means little without them. A surface roughness Ra versus Rz comparison only holds when both are measured with the same settings.
- 1Ra is an averageAll trace points count equally.
- 2Rz is an extreme valueTen points set the result.
- 3Neither describes shapePeak count and slope are invisible to both.
Why one surface gives two numbers that disagree
The ratio between Ra and Rz is not a constant. It is set by how the profile is shaped, and the shape is set by the process that cut it. A sine wave gives an Rz/Ra ratio near 3.2. A turned surface with deep, evenly spaced feed marks usually lands between 5 and 10. Grinding, which leaves a dense and fairly uniform scratch pattern, sits closer to 4 to 6.
This is why converting one to the other with a fixed multiplier is unsafe. A drawing that says Ra 1.6 μm and gets inspected as Rz will pass or fail depending on whether the part was ground or turned. On a rough-turned shaft, Ra 1.6 μm can sit at Rz 10 μm. On a ground one the same Ra may sit at Rz 4 μm.
Honest conversion needs process context. If we know the operation and the expected ratio band, we can predict the partner value within a range. We cannot predict it to a single number, and a supplier who claims otherwise is guessing.
The ratio also drifts across one part. A face milled with a 0.8 mm corner radius can read Ra 0.8 μm and still show Rz 4 to 6 μm near the edge of the cutter path where the stepover marks are deepest.
- 1Sine wave profileRz/Ra near 3.2.
- 2Ground surfaceRz/Ra typically 4 to 6.
- 3Turned surfaceRz/Ra typically 5 to 10.
- 4Few dominant peaksRatio above 10 is possible.
Reading a surface finish callout on a drawing
An ISO 1302 symbol carries the parameter letter and a value, for example Ra 1.6, plus a machining method if the designer wants one. The evaluation length and cutoff are implied by ISO 4288 unless the drawing states them. That default is the source of most arguments between a shop and an inspector.
ISO 4288 picks the cutoff from the expected Ra range. Below Ra 2 μm the cutoff is 0.8 mm. From Ra 2 to 10 μm it is 2.5 mm. Above that it rises to 8 mm. If the drawing says Ra 0.8 and the inspector uses a 2.5 mm cutoff, the number will differ from the supplier's check with no one doing anything wrong.
Specifying Rz needs the same discipline. Rz 6.3 measured over one sampling length is not comparable to a value averaged over five. Write the parameter, the value, the cutoff, and the number of sampling lengths. It takes one line and removes the ambiguity.
For a general machined finish we work to Ra 1.6–3.2 μm. A fine finish call is Ra 0.8–1.6 μm, and Ra 0.2–0.8 μm means grinding, honing, or lapping with dedicated inspection.
- 1State the cutoff0.8 mm, 2.5 mm, or 8 mm.
- 2State the sampling lengthOne or five, not left implicit.
- 3Name the parameterRa and Rz are not interchangeable.
When Ra is the wrong choice and when Rz is
Ra fails on sealing faces. A static O-ring lands on the peaks, not the average. A surface at Ra 0.8 μm with three or four 8 μm peaks will leak, because the rubber bridges the valleys and never seats. The average looked fine and the part still failed the leak test. On hydraulic valve seats and gasket lands, spec Rz, or better the Rk family such as Rpk and Rvk, which describe peak and valley material directly.
Fatigue-critical fillets and radii have the same problem. Cracks start at the deepest valley, and Ra does not see it. Rz or Rmax is the right callout where a stress concentration meets a machined surface.
Rz is the wrong tool for broad appearance requirements. It is computed from a short sampling length, so one anomalous scratch can dominate the result and reject a part that is fine across the whole face. On large castings, weldments, and long shafts the measurement location matters more than the parameter. Ra averaged over several evaluation lengths is more stable there.
Rz is also a poor fit when the requirement is bearing area over a large zone. A plateau surface and a spiky surface can share the same Rz and behave completely differently under load. Use a material ratio parameter instead of forcing Rz to do that job.
- 1Spec Rz for sealsStatic O-ring grooves, valve seats, gasket lands.
- 2Spec Rz for fatigueFillets and radii under cyclic load.
- 3Stay with Ra for appearanceLarge faces, castings, weldments.
- 4Use Rk for plateau surfacesWhen bearing area and oil retention both matter.
Inspection: stylus, cutoff, and filtering
The measurement setup moves the number as much as the cutter does. A 2 μm radius stylus cannot follow valleys narrower than its tip and will under-report Rz on fine ground surfaces. A 5 μm tip is the safer general-purpose choice for shop-floor checks, and it traces deeper valleys more faithfully on turned and milled parts.
Skidded probes are fine for Ra on flat, horizontal surfaces. They are a bad choice for Rz on curved or inclined faces, because the skid follows the form and the roughness trace is distorted. Free-running probes with a proper reference are the correct setup there.
Filtering matters too. The cutoff separates roughness from waviness, and choosing the wrong one shifts the result. Under ISO 4288 the cutoff has to match the expected range. The traverse length should cover at least five sampling lengths so the average means something.
Our inspection runs a raw material check, in-process monitoring, and a final dimensional and finish check before shipment. Reports are available on request. Measurement settings are recorded with the finish result so the number can be reproduced.
- 15 μm tipBetter default for Rz on turned and milled parts.
- 2Free-running probeUse on curved or inclined surfaces.
- 3Five sampling lengthsMinimum traverse for a stable average.
The rule we apply
Specify Ra when the requirement is general finish or appearance over a broad area. Specify Rz when a seal, a gasket land, or a fatigue fillet depends on peak height and valley depth. Always write the cutoff and sampling length next to the value.
Questions we get on finish callouts
Can I convert Ra to Rz with a single multiplier?
No. The ratio depends on the process that made the surface. A ground part may sit near Rz/Ra 5 while a turned part with the same Ra sits near 9. A single multiplier will pass some parts and reject others.
If you need a partner value, give us the operation and we can tell you the expected band, not an exact number.
Which parameter should go on a drawing for a static O-ring groove?
Rz, or the Rk family if plateau behavior matters. A static seal seats on the peaks, so peak height drives leakage. Ra can look acceptable while the peaks are tall enough to bridge the groove.
We normally see Rz 1.6 to 4 μm called out on seal grooves, depending on elastomer hardness and pressure.
Does a 2 μm stylus give the same Rz as a 5 μm stylus?
Not on fine surfaces. A 2 μm tip cannot reach into valleys narrower than its radius, so measured Rz comes out lower than the real value. The gap grows as the surface gets finer.
A 5 μm tip is the practical default for general Rz work. The stylus and its condition should be stated with the result.
Why do my supplier and my lab disagree on the same part?
Usually the cutoff or the number of sampling lengths differs. ISO 4288 ties the cutoff to the expected Ra range, and a mismatch shifts the reading.
Check the traverse length as well. Fewer than five sampling lengths gives a noisy average that moves from spot to spot on the same face.
Is a tighter Ra callout expensive?
It usually adds an operation. Going from Ra 3.2 to Ra 0.8 typically costs a finishing pass, a second setup, or a switch from turning to grinding.
Below Ra 0.4 μm generally needs grinding, honing, or lapping with dedicated fixturing and inspection. That cost lands per part at low volume.
Can you meet Ra 0.2–0.8 μm on aluminum and stainless?
Yes, on the machines and materials we run, with grinding, honing, or lapping as the finishing step. Stainless grades such as 17-4PH hold a fine finish well; soft aluminum needs care to avoid smearing.
Send the drawing with the finish callout and we will confirm the process route in the DFM review.
Send us the drawing with the finish callout
We will review the surface specification, flag any ambiguous callout, and quote with the process route that reaches it.
12-hour quote100% inspection±0.005 mm tolerance