Ra Surface Roughness by CNC Process: Milling, Turning, Grinding and Polishing Bands, Cost Steps and Drawing Callouts

Rule out the callout that costs the most

The most expensive mistake on a finish drawing is a number finer than the function needs. Every step down in Ra is reported to pull in something extra: additional semi-finishing passes, slower feeds, specialised tooling, tighter process control, or a secondary operation such as grinding or polishing^4,6,15,16.

Those extras raise machining time, and machining time is what the part price is built on^4,15. So the cheapest saving available to you is not a cheaper shop — it is not writing the finer number at all.

Design guidance in the cited sources says the same thing from the other direction: roughness specifications should sit within the limits of the primary manufacturing process wherever possible, because very smooth finishes usually require additional processing that adds cost⁶.

One supplier's quoting tool flags what it calls overly restrictive Ra values before production starts, which is a useful habit to copy manually: read your own callout and ask what function actually fails if the surface is one grade rougher¹⁹.

Which process band does your required Ra fall into?

The bands are reported as ordered and largely non-overlapping at the fine end, so the Ra number itself selects the process rather than the other way round¹².

Turning is placed slightly finer than milling at the same nominal capability, which is why a turned bore and a milled face on the same drawing can carry different Ra callouts without contradiction¹².

Treat these as general references, not guaranteed limits — the same source states that actual roughness depends on material, geometry, tooling, machine condition and process parameters¹². A shop that quotes Ra 0.8 µm on a milled face is claiming an optimised-parameter result, not a default one^12,16.

Can the feature come off the primary operation, or does it need a secondary step?

This is the routing decision that changes the quote. Reaching roughly Ra 0.4 µm or finer is reported to require a secondary operation such as grinding or lapping rather than a primary milling or turning pass^{6,13 28} (supplier-reported figure).

Ra 0.8 µm sits on the boundary: one supplier reports hitting it directly off the machine using sharp carbide tooling, a high-RPM spindle and optimised feed rates, without secondary grinding¹⁹, and another reports the same value on aluminium faceplates with 0.5 mm ball end mills at 12,000 RPM under flood coolant¹⁸.

Optimised parameters are the mechanism behind that boundary. Shallow depth of cut and slow feed produce a smoother finish than standard parameters on the same machine⁶, so the difference between Ra 1.6 µm and Ra 0.8 µm off the primary operation is largely cycle time, not a second setup¹² (supplier-reported figure).

Below that, the part leaves the machine. When you route a feature to grinding or lapping, ask the supplier to confirm the achieved Ra on the actual feature rather than the process band, because the sources describe the bands qualitatively and do not tie them to a specific material, hardness or nose radius^12,13.

Writing the Ra callout so the shop cannot substitute a cheaper process

Surface finish on drawings is specified through ISO 1302 or ASME Y14.36M, and the symbol variant carries the material-removal intent^5,26.

A basic check mark allows any method; a circle in the V means removal of material is not permitted, which is how you protect an as-cast or as-forged surface; a bar on the V means material removal is required and an untouched cast surface is not acceptable²⁶.

If you leave the basic symbol on a sealing face, you have not told the shop whether it may grind it. The N-grade system gives you a second anchor.

ISO 1302 defines grades from N1 to N12, each double the previous in Ra value²⁷, and N7 corresponds approximately to Ra 1.6 µm while N8 corresponds approximately to Ra 3.2 µm^{4 12} (supplier-reported figure). Putting the grade next to the Ra value removes the ambiguity about which band you mean.

Note that the N-grade-to-Ra mapping here comes from reference pages, not from the standard text itself^4,27. One more field is worth filling in: the cutoff length.

If no cutoff is specified, 0.8 mm is the default, and that default covers Ra 0.1 µm to 2 µm — most ground, honed and fine-machined surfaces²⁶. For a lapped or polished surface below that range, the drawing should name a shorter cutoff so the inspection instrument measures the right thing²⁶.

Ra alone, or Ra plus Rz?

Ra is the arithmetic mean of the profile heights across the measurement length, which is why it is the most-used parameter on drawings¹⁶. Its weakness is exactly that averaging: a single deep scratch can be averaged out and still pass an Ra check¹⁶.

Rz is the average of the five highest peaks and five deepest valleys in the sampling length, and is reported at roughly 4–5 times the Ra value, so it captures those isolated features¹⁶.

If the surface is a seal face, a bearing bore or anything where one deep valley causes a leak or a stress riser, put both parameters on the drawing and state which one governs acceptance¹⁶.

If the surface is cosmetic or structural, Ra alone is usually enough, and adding Rz only adds inspection cost¹⁶.

Incoming inspection: making the certificate a measured number

Verification is done with surface roughness testers or profilometers, and Gauge R&R studies per AIAG MSA are cited as the practice that makes the measurement data accurate and repeatable^17,18. Neither source quantifies measurement uncertainty, so the method has to be fixed contractually rather than assumed.

• ✓Require the Ra value to be measured with a calibrated profilometer or surface roughness tester, not inferred from the process.
• ✓Ask for the cutoff length and evaluation length used, and confirm they match the drawing callout.
• ✓Ask whether a Gauge R&R study covers the roughness measurement, and request the result.
• ✓For a boundary value such as Ra 0.8 µm, ask for the achieved value on the actual feature, not the process band.
• ✓Confirm the measurement is taken on the specified surface after any secondary operation, not on the pre-finish blank.

When the material is hard, filled or otherwise awkward

The same process does not hold the same Ra across materials. Material grade, filler content and hardness are reported to change machinability and surface behaviour, and the effect is described qualitatively rather than as Ra values per hardness^20,24.

For hardened and case-hardened steel, the reported response is tooling grade: CBN inserts are designed for surface and dimensional tolerance demands at higher cutting parameters^2,21. Tool condition is the other lever, and it is the one you can put in a control plan.

Worn tools are reported to increase surface roughness by 200–300% and to create built-up edge that mars the finish on aluminium¹⁸. Tool runout above 0.002 in TIR is reported to produce chatter marks¹⁸.

Both are process-control items, not drawing items, so they belong in the supplier's control plan alongside the Ra callout^6,18.

If your part is a filled or high-hardness grade, ask the supplier for the achieved Ra on that specific material rather than accepting the generic band, because no source in this set provides an Ra-versus-hardness or Ra-versus-machinability table^20,24.

What to ask before you commit to the process

Two questions the evidence cannot answer must come from the quote: lead time and MOQ for grinding and polishing, and a measured Ra-and-cost example for a comparable feature. No source in this set provides either^12,28.

• ✓Ask for lead time and MOQ for the grinding or polishing step specifically, not for the part as a whole.
• ✓Ask for a measured Ra value and the actual cost delta on a comparable feature from a previous job.
• ✓Ask which process the shop intends to use for each Ra callout, and confirm it matches the band you selected.
• ✓Ask whether the quoted price assumes the Ra is achieved off the primary operation or requires a secondary step.
• ✓Ask for the cost delta between your specified Ra and one grade coarser, so you can see what the finer number is buying.

Where the sources disagree on the Ra 0.8 µm and Ra 0.4 µm boundary

The two disputes are about where the primary operation stops and the secondary step begins. Both sides stay visible here because the answer depends on the shop's tooling and parameters, not on a universal rule^6,13,16,19.

Write the callout, then verify the route in the quote.

What the sources do not establish

• No lead time or MOQ data for polishing or grinding — ask the supplier for these figures directly.
• No measurement uncertainty values, cutoff length or evaluation length for Ra verification — specify the method contractually.
• No Rmax or Rq definitions or relationships — only Ra and Rz are covered in this evidence.
• No Ra-versus-material-hardness or Ra-versus-machinability-rating data table — ask for achieved Ra on the specific material.
• No named case study with measured achieved Ra and actual cost for a specific feature — request a comparable example from the quote.
• No itemized cost breakdown by machining time, tool wear, labour and equipment hourly rate — the cost figures cited are screening estimates, not itemized breakdowns.
• No governing-standard text (ISO 1302, ASME Y14.36M, ISO 21920) reproduced — only secondary descriptions are available.
• No best-case (as opposed to typical) Ra values stated per process with the conditions that enable them — the bands cited are general references, not guaranteed limits.