Every machining process leaves a characteristic roughness, and no amount of care moves a process far outside its band. That is what makes a CNC surface finish chart useful: it tells you which process you are implicitly buying when you write an Ra value on a drawing, and whether that value is reachable in one operation or needs a second.
This page tabulates the achievable Ra band by process, shows how to predict the finish a turning pass will produce before cutting anything, and identifies where a finer number makes the part worse rather than better. How to write the call-out itself is covered in our companion guide to specifying surface finish on a drawing, and applied coatings in our guide to metal surface finishing.
Key takeaways
- Process sets the band. Milling and turning land around Ra 1.6–6.3 µm, grinding around Ra 0.1–1.6 µm, and lapping or polishing below Ra 0.2 µm.
- MW+ works to Ra 3.2 µm as-machined, Ra 0.4 µm fine-machined and Ra 0.1 µm polished. Crossing between those bands means adding an operation, not turning a dial.
- Turned roughness is predictable before you cut: Rz ≈ f² ÷ (8r). Halving the feed cuts theoretical roughness to a quarter — worked below.
- The governing standards have changed. ISO 21920-1, -2 and -3 (2021) supersede ISO 1302, ISO 4287 and ISO 4288, so a drawing should state which convention applies.
- Ra without a cut-off length is incomplete. A λc of 0.8 mm is the usual default in the Ra 0.1–2 µm range, and changing it changes the number.
- Smoother is not automatically better. Oil retention, coating adhesion and plateau-honed bores all need controlled roughness, not the minimum available.
- CNC surface finish chart: Ra by process
- What does Ra actually measure?
- Predicting the Ra a turning pass will give
- Which standard governs a finish call-out?
- Why do two labs report different Ra on the same surface?
- Does the material change the achievable finish?
- Choosing a band by function
- When a finer finish is the wrong specification
- Surface finish at MW+
- Frequently asked questions
CNC surface finish chart: Ra by process
Typical achievable bands, with the legacy ISO N grade that still appears on drawings in circulation. Bands overlap because tooling, rigidity, coolant and material all move the result; treat the table as the range a competent shop can hold, not a guarantee for any single setup.
| Process | Typical Ra (µm) | Ra (µin) | Legacy N grade | Where it is used |
|---|---|---|---|---|
| Sawing, flame cutting | 12.5–25 | 500–1000 | N10–N11 | Stock preparation, non-functional faces |
| Rough milling and turning | 6.3–12.5 | 250–500 | N9–N10 | Material removal before finishing |
| Milling, general finish pass | 1.6–6.3 | 63–250 | N7–N9 | Housings, brackets, general machined faces |
| Turning, general finish pass | 0.8–6.3 | 32–250 | N6–N9 | Shafts, spigots, turned bodies |
| Drilling | 1.6–6.3 | 63–250 | N7–N9 | Clearance and tapped holes |
| Reaming | 0.8–1.6 | 32–63 | N6–N7 | Dowel and locating bores |
| Boring | 0.4–3.2 | 16–125 | N5–N8 | Bearing housings, large bores |
| Wire EDM | 0.4–3.2 | 16–125 | N5–N8 | Profiles in hardened material |
| Sinker EDM | 0.8–6.3 | 32–250 | N6–N9 | Cavities, sharp internal corners |
| Grinding | 0.1–1.6 | 4–63 | N3–N7 | Hardened shafts, seal journals, gauge faces |
| Honing | 0.05–0.8 | 2–32 | N2–N6 | Hydraulic bores, cylinder liners |
| Lapping | 0.012–0.4 | 0.5–16 | N1–N5 | Sealing faces, optical and metrology surfaces |
| Polishing, superfinishing | 0.012–0.2 | 0.5–8 | N1–N4 | Bearing races, medical contact surfaces |
| Bead blasting | 1.6–6.3 | 63–250 | N7–N9 | Uniform matte appearance, coating preparation |
The practical reading is where the bands stop overlapping. Anything at or above Ra 1.6 µm is reachable by CNC milling services or CNC turning services in the same setup as the dimensional work. Below roughly Ra 0.4 µm a second operation — grinding, honing, lapping or polishing — becomes necessary, and that is the point at which the finish call-out starts to drive the process route rather than follow it.
What does Ra actually measure?
Ra is the arithmetic mean of the absolute deviations of the measured profile from its mean line, over a defined evaluation length. It is an average, which is its strength and its weakness: it is stable and repeatable, and it is nearly blind to isolated features.
Two surfaces can share an Ra value and behave completely differently. A ground surface with uniform fine scratches and a surface carrying a few deep tool marks can average to the same number, yet only one of them will seal. That is why parameters beyond Ra exist.
- Rz — mean of the largest peak-to-valley heights in each sampling length. Sensitive to the outliers Ra averages away. Roughly four times Ra on a turned surface, but the ratio is not fixed and should never be used as a conversion on a controlled feature.
- Rt — total height of the profile over the whole evaluation length. The single worst feature.
- Rsk — skewness. Negative skew means plateaus with valleys, which is what a plateau-honed bore needs for oil retention. Ra cannot see this at all.
- Rsm — mean width of the profile elements, which describes spacing rather than height and matters for coating adhesion and for optical appearance.
All of these are defined in ISO 21920-2:2021, which supersedes the withdrawn ISO 4287, and in ASME B46.1.
Predicting the Ra a turning pass will give
On a turned surface the dominant roughness is geometric: the tool nose traces a helix, leaving scallops whose height is set by feed per revolution and nose radius. The theoretical peak-to-valley height is:
Rz ≈ f² ÷ (8 × r), with f the feed in mm/rev and r the nose radius in mm.
Take f = 0.20 mm/rev and r = 0.8 mm. Then Rz ≈ 0.20² ÷ (8 × 0.8) = 0.04 ÷ 6.4 = 0.00625 mm, which is 6.25 µm. Using the rough Ra ≈ Rz ÷ 4 relationship for a turned profile, that is about Ra 1.6 µm. Repeating the arithmetic across a range:
| Feed f (mm/rev) | Nose radius r (mm) | Rz = f² ÷ (8r) | Indicative Ra | Change vs the first row |
|---|---|---|---|---|
| 0.20 | 0.8 | 6.25 µm | ≈ 1.6 µm | baseline |
| 0.10 | 0.8 | 1.56 µm | ≈ 0.39 µm | −75% |
| 0.05 | 0.8 | 0.39 µm | ≈ 0.10 µm | −94% |
| 0.10 | 0.4 | 3.13 µm | ≈ 0.78 µm | −50% |
| 0.10 | 1.2 | 1.04 µm | − | −83% |
Two things fall out. Because feed appears squared, halving it cuts theoretical roughness to a quarter — and halves the metal removal rate, which is the real price of a finer finish. And a larger nose radius improves finish at the same feed, which is why a finishing insert is usually the cheaper lever before the feed is touched. The calculation is a floor, not a promise: it predicts the best a geometrically perfect setup could do.
Which standard governs a finish call-out?
This is the single most common out-of-date item on machining drawings. The ISO surface texture standards were restructured in 2021, and the documents most drawings still name have been withdrawn.
| Current standard | Covers | Supersedes |
|---|---|---|
| ISO 21920-1:2021 | Indication of surface texture on drawings | ISO 1302:2002 (withdrawn) |
| ISO 21920-2:2021 | Terms, definitions and parameters including Ra, Rz, Rsk | ISO 4287:1997 (withdrawn) |
| ISO 21920-3:2021 | Specification operators: cut-off, evaluation length, filters | ISO 4288:1996 (withdrawn) |
| ASME B46.1 | North American surface texture standard | Current; used in place of the ISO set |
Naming a withdrawn standard is not itself an error — a legacy drawing frozen under change control may have to keep it. What causes disputes is naming neither, or mixing an ISO symbol with an ASME default. State the standard and edition next to the call-out and the ambiguity disappears.
Why do two labs report different Ra on the same surface?
Almost always because they used different filtering. Ra is not a property of the surface alone; it is a property of the surface as seen through a specified cut-off. Waviness longer than the cut-off wavelength λc is filtered out, so a longer cut-off admits more long-wave content and raises the reported number.
| Ra range (µm) | Cut-off λc (mm) | Evaluation length (mm) |
|---|---|---|
| 0.006 to 0.02 | 0.08 | 0.4 |
| Over 0.02 to 0.1 | 0.25 | 1.25 |
| Over 0.1 to 2 | 0.8 | 4 |
| Over 2 to 10 | 2.5 | 12.5 |
| Over 10 to 80 | 8 | 40 |
Three other things move the reading. Stylus tip radius: a larger tip bridges narrow valleys and reports a smoother surface. Measurement direction: roughness measured along the lay reads lower than across it, so the direction must be specified on any surface with a strong lay. And location: a finish varies across a face, so a single trace at the operator’s choice of spot is not evidence about the whole feature. The report should state cut-off, direction and location, and the practice behind that sits under CNC machining quality control.
Does the material change the achievable finish?
Substantially, because the limiting factor at fine finishes is how the material forms a chip rather than the geometry of the tool path. Free-cutting grades shear cleanly and reach the bottom of the process band; ductile and gummy grades tear and smear, and need either a different tool geometry or a secondary operation to get there.
| Material | Behaviour at the cutting edge | Practical route to Ra 0.4 µm or better |
|---|---|---|
| Aluminium 6061-T6 | Clean short chips, low cutting force | Fine machining, often without a secondary operation |
| Brass C36000 | Free-breaking chips, very stable | Fine turning direct from the machine |
| Stainless 303 | Free-machining additions break the chip | Fine machining; grinding for the tightest bands |
| Stainless 304 / 316L | Work-hardens ahead of the tool, stringy chips | Fine machining plus grinding or polishing |
| Alloy steel 4140 | Abrasive but predictable; hardens well | Grinding after hardening |
| Titanium Ti-6Al-4V | Heat concentrates at the edge, springback on thin walls | Fine machining plus polishing |
| Copper C101, acetal | Gummy; smears rather than shears | Sharp uncoated or diamond tooling, then polishing |
Choosing a band by function
Work backwards from what the surface has to do. The bands below are the usual engineering answers, and each one should be checked against the specific application rather than adopted wholesale.
- Non-functional and general faces: Ra 3.2–6.3 µm. The as-machined condition. Specifying anything finer here adds time for no return.
- Bolted joints and mating faces: Ra 1.6–3.2 µm. Enough flatness and contact area without a finishing operation.
- Static seals, O-ring grooves: Ra 0.8–1.6 µm on the groove, finer on the sealing face. Peaks matter more than the average, so consider adding an Rz limit.
- Dynamic seals and sliding surfaces: Ra 0.2–0.8 µm. Too rough abrades the seal; too smooth starves the interface of lubricant.
- Bearing journals and precision fits: Ra 0.1–0.4 µm, normally ground.
- Optical, metrology and medical contact surfaces: Ra 0.05 µm and below, by lapping or polishing, with the measurement method agreed in advance.
When a finer finish is the wrong specification
Lubricated sliding interfaces
A surface that is too smooth cannot hold a lubricant film. This is why engine cylinder bores are plateau-honed rather than polished: the process leaves flat load-bearing plateaus with a controlled valley structure underneath. An Ra limit alone will not produce that; the drawing needs a skewness or bearing-ratio requirement as well.
Surfaces about to be coated
Thermal spray and many adhesive systems need mechanical keying, so an over-finished substrate adheres worse than a controlled rough one. Specify the pre-treatment condition explicitly, and state whether the Ra requirement applies before or after coating — a coating changes the measured value, and a drawing that is silent on this creates an argument at incoming inspection.
When the tolerance cannot support it
A grinding or lapping operation removes material, so a fine finish and a tight size tolerance have to be planned together. Adding a finishing operation to a part already at its lower limit means scrapping it. Say which requirement dominates if they conflict.
When the finish cannot be verified
A deep narrow slot or a small internal bore may be physically inaccessible to a stylus. Specifying Ra 0.2 µm in a place no instrument can reach produces a requirement that can be neither met nor disproved. Ask whether the feature is measurable before the drawing is released.
Surface finish at MW+
MW+ has machined precision components in Shenzhen since 2015, from a 15,000 m² facility running 60+ CNC machining centres with 120+ engineering and quality professionals and 70+ materials in regular use. Over a million parts have shipped to 50+ countries at a 99% on-time delivery record.
The standard finish ladder is Ra 3.2 µm as-machined, Ra 0.4 µm fine-machined and Ra 0.1 µm polished, alongside a tolerance system of ±0.01 mm general to ISO 2768-m, ±0.005 mm precision and ±0.001 mm on selected features. Finish requirements are reviewed against the process plan at quotation, so a call-out that needs a secondary operation is identified before the order rather than after it. Every order ships with a certificate of conformance, a CMM inspection report and material certificates, with the achieved finish stated where it is a controlled characteristic; the wider inspection method is described in our guide to the CMM inspection process.
Work runs under ISO 9001:2015, AS9100D, ISO 13485, IATF 16949 and NADCAP accreditation. Finished families are catalogued as CNC precision parts, hardened profiles are cut by wire EDM services, and the full process list sits under CNC machining services.
Frequently asked questions
What Ra should I put on a drawing if I have no particular requirement?
Use the as-machined condition, around Ra 3.2 µm, as a general note in the title block, and call out finer values only on the specific faces that need them. A blanket fine finish across a whole part is the most expensive way to say nothing, because it forces every surface through a finishing route including the ones no one will ever touch.
Can I convert Rz to Ra, or the other way round?
Not reliably. The ratio depends on the process and typically runs somewhere between about four and seven on machined surfaces, but it is a description of a profile shape rather than a constant. If a customer specifies Rz, measure Rz. Converting is acceptable for a rough sanity check and unacceptable as conformity evidence.
Why did my part measure inside the Ra limit on your report and outside it on mine?
Compare the cut-off length, the measurement direction relative to the lay, the stylus tip radius and the exact location traced before anything else. Any one of those will move the number, and a difference in two of them can easily account for a disagreement that looks like a process problem. Agreeing those four items in advance removes most surface finish disputes entirely.
Does a finer Ra improve fatigue life?
It can, because surface features act as stress concentrations and cracks generally start at the surface. But the relationship depends on material, stress level and loading mode, so it is a reason to engage a materials engineer rather than a reason to halve the number on the drawing. In highly loaded parts, residual compressive stress from shot peening often matters more than roughness alone.
How do I specify a finish on a surface that will be anodised or plated?
State whether the requirement applies before or after the treatment, and state the treatment thickness, because the coating changes both the dimension and the texture. On decorative or matte requirements, specify the pre-treatment condition — a bead-blasted substrate and a fine-machined one look entirely different after the same anodising process.
Is a finish achievable on every material at the same cost in time?
No. Ra 0.8 µm in free-cutting brass is routine from the finishing pass, while the same figure in 316L or titanium usually means an extra operation. That difference shows up as machine hours, so if two candidate materials both satisfy the function, ask for both to be quoted with the finish requirement attached.
My drawing still says ISO 1302 and ISO 4287. Is it invalid?
No. Those documents are withdrawn but remain the governing convention for drawings issued under them, and a great many controlled drawings still name them legitimately. The risk is mixing conventions across a revision. If you are reissuing the drawing, move to the ISO 21920 series; if you are not, leave it alone and make sure the supplier knows which convention they are inspecting to.
