Surface Finish Requirements for Tight-Tolerance CNC Machining: Ra, Rz, and What Aerospace & Defense Buyers Actually Need to Specify

At a Glance

• Ra is the most-specified surface texture parameter, but it hides isolated peaks — for sealing surfaces you also need Rz
• Ra 125 µin (3.2 µm) is the baseline for aerospace fatigue-life analysis; over-specifying below Ra 32 µin (0.8 µm) is a top-three RFQ cost driver
• ISO 21920-2:2021 replaces ISO 4287 — same measurement, sometimes different reported value; always confirm which standard the drawing invokes
• Olympus verifies Ra/Rz per ISO 21920-2 and ASME B46.1-2019 with NIST-traceable comparator specimens
• 7-point buyer checklist and 6-question FAQ at the bottom of this post

A ±0.0005-inch tolerance means nothing if the surface finish is wrong. A dimensionally perfect sealing face that leaks, a landing-gear pin that galls on install, a titanium airfoil that fails fatigue testing at half its design cycles — every one of these is a surface-finish failure dressed up as a dimensional problem.

Surface finish is where drawings, inspection, and function collide. Buyers over-specify Ra and pay for it. They under-specify Rz and parts leak, gall, or fail fatigue. This post is what an aerospace or defense buyer actually needs to know before RFQ — the parameters, the standards, the typical values by process, and the mistakes that cost programs weeks of rework.

Ra vs Rz vs Rq: What Each Parameter Actually Measures

Surface texture is a filtered profile. A stylus (or optical probe) traces the machined face, a Gaussian filter separates roughness from waviness, and the remaining profile is characterized by a family of Rx parameters. Three matter for 95% of CNC drawings:

• Ra — arithmetic mean roughness. Average absolute deviation from the mean line, integrated over the evaluation length. The most-specified parameter on drawings because it is stable, repeatable, and cheap to measure. Its weakness: Ra hides isolated peaks and valleys. A profile with one 5 µm valley and a mostly-smooth surface can report the same Ra as a uniformly rougher surface.
• Rz — maximum height of the profile. The sum of the largest peak and the largest valley within the sampling length (or the maximum peak-to-valley height within the evaluation length under ISO 21920-2). Critical for sealing surfaces: a single 8 µm peak will bridge an O-ring even if Ra is well within spec.
• Rq (RMS) — root-mean-square deviation. More sensitive to occasional large deviations than Ra. Rq ≈ 1.11 × Ra for typical machined surfaces; Rq is preferred in optics and semiconductor tooling.

Typical Rz/Ra ratios by process (useful when a drawing calls out only one and you need to estimate the other): turned ~4–5, milled ~4–6, ground ~5–7, honed ~5–7, lapped ~6–8, EDM ~5–6. These are conversion estimates only — never substitute a calculated Rz for a measured one on a critical feature.

Definitions and math are formalized in ASME B46.1-2019 and ISO 21920-2:2021 (which supersedes ISO 4287).

The 2021 Standards Shift: ISO 21920-2 Replaces ISO 4287

Aerospace primes are updating drawings from ISO 4287 to ISO 21920-2:2021, and if you receive a drawing that cites the old standard on a new revision it is worth a call to the customer before the RFQ closes. The ANSI summary of ASME B46.1-2019 walks through the parallel U.S. update.

Practical differences that show up in inspection reports:

• Rz redefined. Under ISO 4287, Rz was the "mean roughness depth" — averaged across sampling lengths. Under ISO 21920-2, Rz is the maximum peak-to-valley height across the entire evaluation length. On a well-machined surface the two are close; on a surface with one outlier peak they can differ by 20–40%.
• Evaluation length by default. Most parameters are now defined on the evaluation length rather than averaged across sampling lengths. Rp, Rv, and Rz remain peak-to-valley parameters.
• Setting Classes (Sc1–Sc5). Standardized filter/section-length defaults replace the "pick your cutoff" ambiguity that plagued ISO 4287 reports. A drawing that specifies Sc3 leaves no room for interpretation on λc, evaluation length, or stylus tip radius.

Bottom line: the measurement did not change, but the reported value can. Always confirm which standard the drawing invokes — and if in doubt, ask for both.

Typical Ra Values by CNC Process

Note: Ra 125 µin (3.2 µm) is the baseline for aerospace and automotive fatigue-life testing per common industry practice. Over-specifying below Ra 32 µin (0.8 µm) without functional justification is one of the top three cost drivers on aerospace RFQs. See the ASME summary sheet in the B46.1-2019 poster.

Aerospace & Defense Surface Finish Callouts You'll See in the Field

• Turbine blade airfoils: Ra 0.2–0.4 µm, typically via isotropic superfinishing (ISF) after five-axis milling
• Engine housings: Ra 0.8–1.2 µm on mating flanges and bearing bores
• Landing gear: Ra 0.4–0.6 µm on chrome-plated surfaces, preceded by shot peening and polishing
• Fuel system sealing surfaces: Ra 0.3–0.5 µm with a paired Rz callout (typical Rz ≤ 3.2 µm)
• Structural aluminum wing components: Ra 0.6–1.0 µm on datum surfaces to control fatigue crack initiation
• Prime-specific callouts: Boeing D6-56000, GE P1TF14, and Lockheed process specs supersede ASME/ISO defaults — read the specification, not just the drawing note

For controlled defense technical data, all of this work runs inside our ITAR envelope. Olympus Machining is ITAR-registered and CMMC Level 1 (FAR 52.204-21 compliant) — surface finish drawings on ITAR programs are handled the same as any other controlled technical data.

When Ra Isn't Enough: Rz, Rp, Rv, and Lay

Ra is a summary statistic. On non-critical surfaces it is sufficient. On functional surfaces the drawing should specify at least one additional parameter:

• Sealing surfaces — always specify BOTH Ra and Rz. A part that passes Ra can still fail on a single deep valley that ruptures the seal contact.
• Sliding and bearing surfaces — Rp (maximum peak height) matters more than Ra. A high Rp with low Ra means a smooth surface with a few tall asperities that will gall on assembly.
• Lay direction — six standard symbols (=, ⊥, X, M, C, R) per ASME Y14.36. Critical for O-ring seats, where circumferential lay (C) is required so the seal does not spiral into the machining pattern.
• Skewness (Rsk) and kurtosis (Rku) — Rsk < 0 indicates a surface dominated by valleys (good for oil retention); Rku describes the peakedness of the height distribution. Used with Abbott-Firestone (Rk, Rpk, Rvk) analysis on plateau-honed bearing surfaces.

How Olympus Verifies Surface Finish

Every surface texture callout on a customer drawing is captured with instrument S/N, cutoff length (λc), evaluation length, and measurement direction relative to lay. Our verification chain:

• Contact profilometer (diamond stylus) for Ra, Rz, Rq, Rp, Rv per ISO 21920-2 and ASME B46.1-2019. Stylus tip radius, cutoff, and evaluation length are recorded on every report.
• Cross-referenced with CMM measurement on our Haas HMM 430 (milled parts) and Chien Wei CWB-450-CNC (turned parts) for form/waviness separation — see our CMM inspection capabilities post for the full metrology chain.
• PPAP/ISIR-compatible reports — every finish callout appears on its own row with measurement result, instrument S/N, and pass/fail.
• AS9102 Form 3 characteristic accountability — surface texture callouts are balloon-numbered and reported alongside dimensional characteristics. See our AS9102 first article inspection checklist.
• Traceability to NIST via calibrated roughness comparator specimens and annual instrument calibration.

7-Point Buyer Checklist: Specifying Surface Finish on Your Drawing

1. Set a general note baseline. Example: "Unless otherwise stated, all machined surfaces Ra 3.2 µm (125 µin) max." This one line eliminates most RFQ back-and-forth on non-critical features.
2. Call out tighter Ra only on functional surfaces. Every Ra callout below the general note adds cost. Justify each one against a sealing, sliding, fatigue, or cosmetic function.
3. For seals and hydraulic surfaces, specify BOTH Ra and Rz. Ra alone will let a peak-heavy surface pass inspection and fail in the field.
4. Specify lay direction for O-ring seats and dynamic seals. Use the ASME Y14.36 symbols (=, ⊥, X, M, C, R). Circumferential lay (C) is the default for round seal grooves.
5. Reference the standard. "Per ISO 21920-2:2021" or "Per ASME B46.1-2019" on the drawing removes ambiguity about filter, cutoff, and evaluation length.
6. Flag measurement certification in the RFQ. A signed surface finish report per feature adds inspection time — note it up front so the quote reflects it.
7. Avoid over-specifying Ra < 0.8 µm without functional justification. Every step tighter than N6 typically doubles the finishing cost and adds a secondary process (grinding, honing, superfinishing).

Common Surface Finish Mistakes That Kill Aerospace RFQs

• No surface finish callout at all. The drawing defaults to "as-machined," which every shop interprets differently. Ambiguity becomes an NCR after the first shipment.
• Ra only on a seal. A dynamic seal that requires Rz control will pass Ra and leak in service.
• Wrong standard cited. A 2024 drawing that cites ISO 4287 forces the shop to guess whether the customer wants old-Rz or new-Rz. Cite ISO 21920-2 explicitly.
• No lay specification on a rotary seal. Axial lay on a rotating O-ring groove is a certified leak path.
• Ra < 0.4 µm on a non-critical surface. Doubles the machining cost, adds a secondary process, and rarely adds any function.

Frequently Asked Questions

Need certified surface-finish reports on your aerospace or defense parts?

Olympus Machining provides Ra/Rz measurement per ISO 21920-2 and ASME B46.1-2019 alongside AS9102 Form 3 characteristic accountability on every shipment for Aerospace & Defense programs.

Get a quote AS9102 first article inspection checklist