AS568 · ISO 3601 · MIL-PRF cross-references, 15 compounds

Surface Finish for O-Ring Grooves: Ra, Rz and the 8–16 µin Rule

Roughness decides whether a seal conforms or leaks along the machining marks. What Ra and Rz actually measure, target values for static and dynamic grooves, and why too smooth is also a failure.

9 min read 1,163 words Updated 2026-11-17

An O-ring seals by filling the microscopic valleys in the surface it is pressed against. If the valleys are deeper than the rubber can reach at the available contact stress, fluid passes through them, and no amount of tightening will stop it. Surface finish is therefore not cosmetic: it is part of the sealing mechanism.

It is also the specification most often copied from a drawing without understanding, which is how joints end up with a 63 µin finish that "looks smooth" and leaks at 50 psi.

A row of O-rings on a machined surface
The valleys, not the average, are what the rubber has to fill.

Ra and Rz are not the same measurement

Ra is the arithmetic average of the profile height deviations from the mean line, over the sampling length. It is an average, it smooths out the extremes, and it is the number most drawings specify.

Rz is the average of the maximum peak-to-valley heights within each sampling length. It describes the deepest features, which are the ones a seal has to bridge.

The two are related but not interchangeable: for a typical turned or ground surface, Rz is roughly 4 to 6 times Ra. A surface specified as Ra 0.4 µm has an Rz of about 1.6–2.4 µm, meaning individual valleys are 1.6–2.4 µm deep even though the average is 0.4.

This matters because two surfaces with the same Ra can seal very differently. A surface with many shallow scratches and a surface with a few deep grooves average out to the same Ra; the second leaks.

Specify Rz where the seal contacts the surface.

Ra tells you about the general quality of the machining; Rz tells you about the deepest valleys the rubber has to fill. For static seals, an Rz limit alongside the Ra target catches the surface that averages well and seals badly.

Target values

ContactRa (µin)Ra (µm)Rz (µm)
Static groove floor8–160.2–0.41.0–2.5
Static groove side walls16–320.4–0.82.5–5.0
Dynamic sealing surface (rod or bore)4–80.1–0.20.6–1.5
Static mating face (flange)16–320.4–0.82.5–5.0
Non-sealing surfaces63–1251.6–3.2—

Two notes on reading this table. First, the dynamic values are tighter than the static ones because a dynamic seal is being rubbed, and a rough surface abrades it. Second, the table describes the surface the rubber touches, not the whole part: a groove floor at 8 µin with side walls at 63 µin is a normal, sensible specification.

Why too smooth is also a failure

Below about 2 µin Ra (0.05 µm), the surface holds no lubricant film. In dynamic service, rubber in direct contact with polished metal has high friction, and high friction produces stick-slip: the seal moves in jerks rather than sliding, wears locally, and generates heat.

In static service, a very smooth surface can also be worse than a moderately smooth one, because there is nothing for the rubber to key into and the seal is more sensitive to any contamination or fluid film at the interface.

The practical band is therefore a band, not a target: 8–16 µin for static, 4–8 µin for dynamic. Polishing beyond the low end buys nothing and costs money.

Direction matters more than the number

A turned groove has circumferential marks, a ground surface has a directional pattern, and a milled surface has a path. The seal only has to bridge the marks that cross its sealing path:

  • Circumferential marks in a radial groove floor run parallel to the seal. They are tolerable, because the rubber does not have to bridge them across the leak path.
  • Axial scratches run across the seal and form a direct leak channel. Even a single deep axial scratch can leak a joint that meets its Ra specification.
  • Spiral marks from a single-point threading tool are the worst case, because they form a continuous helical path under the seal. This is the classic cause of a shaft seal that leaks no matter how many times it is replaced.

A finish specified only as "Ra 16 max" does not capture any of this. Add "no circumferential helical marks; axial scratches not permitted in the sealing band".

Contamination and finish are the same problem

A 16 µin surface with metal chips in the valleys behaves like a much rougher surface. Cleaning before assembly is part of the finish specification, and a joint that is machined to 8 µin and assembled dirty will fail like one machined to 63 µin.

The same applies to coatings and plating: a chromed rod with a porous deposit has a roughness the drawing does not describe, and a phosphate coating on a cover changes the effective surface the seal sees.

Measuring it

Three methods, in increasing order of rigour:

  1. Comparator plates. A visual and tactile comparison with a standard set. Adequate for shop-floor acceptance of non-critical surfaces, and useless for a 4–8 µin requirement.
  2. Stylus profilometer. The standard method. Measure in at least two directions, because a directional surface will read differently along and across the machining marks.
  3. Casting or optical methods for surfaces the stylus cannot reach, such as the bottom of a narrow groove.

Record the direction of measurement with the result. An Ra value without a direction is half a measurement on a turned part.

What rough surfaces do in service

The failure signature of a too-rough surface is a joint that weeps at low pressure and holds at high — the same signature as insufficient squeeze, which is why finish problems are often misdiagnosed as gland problems. At low pressure there is not enough contact stress to push the rubber into the valleys; at high pressure the fluid load closes the gap.

In dynamic service, a rough surface produces a different signature: a polished band on the seal where it has been abraded, and fine rubber debris in the fluid.

Frequently asked questions

What surface finish should an O-ring groove have?

8–16 µin Ra (0.2–0.4 µm) on the groove floor for static seals, 4–8 µin (0.1–0.2 µm) on dynamic sealing surfaces. Side walls can be rougher, typically 32 µin.

Is a smoother surface always better for sealing?

No. Below about 2 µin the surface holds no lubricant film, friction rises and dynamic seals stick-slip. The target is a band, not a minimum.

Why does my joint leak when the finish meets the drawing?

Most often because the specification is Ra only and the surface has deep valleys or a helical mark that Ra averages away. Add an Rz limit and a direction requirement.

Can I improve the finish after machining?

Yes, by lapping, honing or polishing, but each process has its own direction and its own risk of rounding edges. Re-measure after any post-processing, and check that the groove depth is still in tolerance.

Does plating affect the finish requirement?

Yes. A plated or coated surface has its own roughness, often higher than the base metal, and a porous deposit can trap fluid. Specify the finish on the finished surface, not on the machined one.

Parts and data mentioned above
Scope

Figures quoted here come from published standards and compound datasheets. They describe typical behaviour, not a guarantee for a specific batch or application. Prototype and life-test before release.