Surface Finish Specification and Measurement: Ra, Rz, and Rmr Explained

Surface finish failures in CNC machining are more often a specification or measurement problem than a machining problem. A machinist who hits Ra 0.8 um on the profilometer can still fail inspection if the cutoff length, evaluation length, or measurement direction is wrong. This guide covers how to correctly specify, measure, and verify Ra, Rz, and Rmr so that parts pass inspection the first time.

Quick Surface Finish Troubleshooting Reference

Surface Finish Problem	Primary Action	Expected Ra Impact
Ra reading 30-50% lower at supplier than at customer	Match cutoff length to ISO 4288 (0.8 mm cutoff for Ra 0.1-2.0 um) on both instruments	Aligns measurements within ~10% by removing the dominant filter-bandwidth error
Specified Ra met but seal still leaks	Add Rz max callout per ISO 4287 (Rz/Ra typically 4-7×)	Catches isolated deep valleys that average Ra hides; eliminates single-valley leak paths
Ra reads 2-5× higher than expected	Re-measure perpendicular to lay direction per ISO 4288	Restores correct Ra reading; lay direction is the largest single setup variable
Bearing surface wears too fast after break-in	Specify Rmr(c) = 70-90% at 10% of Rz cutting depth	Increases load-bearing area; plateau profile carries load while valleys retain oil
Profilometer reads inconsistently across same surface	Verify stylus tip condition with reference specimen monthly	Detects chipped or contaminated 2/5 um tips that under-report Ra by 10-30%
Ra value drifts during measurement	Allow 30-60 min thermal stabilization before measuring	Eliminates thermal-expansion artifacts at the +/-0.001 mm level
Ra-only callout permits over-finishing	Add lower Ra limit (e.g., Ra 0.4 um minimum) for break-in surfaces	Preserves micro-asperities needed for oil retention during running-in

Ra, Rz, Rmr: What Each Parameter Actually Measures

Ra averages everything while Rz captures the worst single defect, so two surfaces with identical Ra 0.8 um can leak or seal depending on which parameter you actually controlled.

Ra (arithmetic average roughness) is the mean absolute deviation from the profile centerline over the evaluation length. It is the most commonly specified parameter because it is easy to measure and universally supported by instruments. However, Ra averages everything -- a surface with deep isolated scratches and a surface with uniform shallow peaks can produce identical Ra readings while performing very differently under load or with seals.

Rz (maximum height of the profile) is defined in ISO 4287 as Rp + Rv: the maximum peak height plus the maximum valley depth within a single sampling length. Rz is sensitive to isolated defects that Ra masks. For sealing surfaces, Rz matters more than Ra because a single deep valley can create a leak path even when the average roughness is within tolerance. Note: older drawings using DIN 4768 define Rz differently — as the mean of five peak-to-valley heights across the evaluation length — which yields a lower value than ISO 4287 Rz for the same surface.

Rmr (material ratio / Abbott-Firestone curve) describes what percentage of the surface lies above a given depth. At a cutting depth of 10% of Rz, the material ratio tells you how much of the surface will carry load after initial break-in. A bearing surface with Rmr(c) = 80% at 10% depth has far more load-bearing area than one with Rmr(c) = 40%, even if both share the same Ra.

Parameter	What It Measures	Best For	Limitation
Ra	Average roughness	General-purpose specification	Hides isolated defects
Rz	Peak-to-valley height	Sealing surfaces, coating adhesion	Sensitive to single outliers
Rmr	Material ratio at depth	Bearing surfaces, wear resistance	Requires profile analysis software

✓

Two Surfaces, Same Ra, Different Function

Two turned surfaces can both measure Ra 0.8 um yet behave completely differently. One has uniform tool marks with consistent peak spacing. The other has deep feed marks with plateaued peaks from honing. The first fails as a seal surface because the deep valleys leak. The second performs well because Rz is low and Rmr is high. Specifying only Ra gives the machinist a target but does not guarantee functional performance.

Specifying Surface Finish on Engineering Drawings (ISO 1302)

ISO 1302 is the indication standard for surface texture on technical drawings, and incomplete callouts are the upstream cause of most supplier-customer disputes downstream at inspection. ISO 1302 defines the standard symbol for surface texture annotation. Incorrect or incomplete callouts cause disputes between machining and inspection.

The basic symbol is a checkmark shape with extensions for each parameter. The full annotation includes:

Upper limit -- the maximum allowed roughness (e.g., Ra 1.6 um)
Lower limit -- the minimum roughness, when specified (e.g., Ra 0.4 um minimum for bearing break-in)
Manufacturing process -- placed above the symbol when a specific process is required (e.g., "ground", "turned", "lapped")
Lay direction -- the symbol indicating the predominant pattern direction: = parallel to projection plane, perpendicular, X crossed, M multi-directional, C circular, R radial
Machining allowance -- material to be removed, in millimeters
Evaluation length and cutoff -- specified when the default per ISO 4288 is not appropriate

Common specification mistakes:

Calling out Ra without specifying the evaluation length -- the default per ISO 4288 may not match the inspector's assumption
Omitting lay direction on sealing surfaces where the leak path depends on the scratch pattern relative to the seal
Specifying a single upper limit when both upper and lower limits are functionally necessary (over-polished bearing surfaces skip the break-in phase)
Using N-grade callouts (N6, N7) without the corresponding Ra value -- ambiguous across older and current standard editions

An Ra-only callout with no lower limit is the most common drawing error for sliding-contact and bearing surfaces — it permits mirror-polished finishes that carry no oil and fail under initial load because the micro-asperities needed for running-in have been removed.

Profilometer Setup: Cutoff, Evaluation Length, and Filters

Wrong cutoff length per ISO 4288 shifts Ra readings by 20-50% even when the surface itself is unchanged, making it the single largest source of supplier-customer measurement disputes. Incorrect profilometer setup is the single most common cause of measurement disputes between suppliers and customers. ISO 4288 defines the cutoff length based on the expected Ra range, and getting this wrong shifts the reading by 20-50%.

Cutoff length selection per ISO 4288:

Expected Ra Range (um)	Cutoff Length (mm)	Evaluation Length (mm)
0.006 - 0.02	0.08	0.4
0.02 - 0.1	0.25	1.25
0.1 - 2.0	0.80	4.0
2.0 - 10.0	2.50	12.5
10.0 - 80.0	8.00	40.0

The evaluation length equals 5x the cutoff length. This provides five sampling lengths for statistical reliability. Measuring with fewer than five cutoffs reduces confidence and is non-compliant with ISO 4287.

Filter selection (ISO 16610): ISO 16610 is the current standard family for surface profile filtration and is preferred over older 2RC filtering because Gaussian filters produce readings 5-15% lower than 2RC on the same surface, eliminating a hidden bias. The Gaussian filter is the current ISO standard and should be used unless the drawing specifies otherwise. Older instruments may default to the 2RC filter, which produces readings 5-15% higher than the Gaussian filter on the same surface. If supplier and customer use different filter types, readings will disagree even with identical setups. Confirm which filter is active and document it on the inspection report — this one setting accounts for much of the systematic inter-lab disagreement on conforming surfaces. Profilometer hardware itself (stylus tip radius, traverse speed, vertical resolution) is governed by ISO 3274, which is the contact stylus instrument specification that ISO 4287 measurements rely on.

⚠

Wrong Cutoff Is the Most Common Measurement Error

Using a 0.25mm cutoff when the surface is Ra 1.2 um (which requires 0.8mm per ISO 4288) filters out the actual roughness profile and returns a reading 30-50% lower than the true value. Parts pass inspection at the supplier and fail at the customer because of this single setup difference. Select cutoff based on the expected Ra range before measuring — this is the first setting to verify whenever supplier and customer readings disagree.

Common Measurement Errors That Cause False Failures

Measuring perpendicular to the lay direction is mandated by ISO 4288 because parallel-to-lay measurements under-report Ra by 2-5×, producing parts that pass at the supplier and fail at the customer. Beyond cutoff selection, several practical errors cause parts to fail inspection despite meeting the actual surface requirement.

Measuring across the lay direction. Ra measured perpendicular to the machining marks typically reads 2-5x higher than Ra measured parallel to the lay. ISO 4288 specifies measurement perpendicular to the lay as the default, but many operators measure along the convenient direction without checking.
Worn or contaminated stylus tip. A standard diamond stylus has a 2 um or 5 um tip radius. A chipped or contaminated tip cannot track fine surface features, producing artificially low readings. Verify tip condition monthly using a reference specimen.
Thermal expansion during measurement. Moving a part from a cold warehouse to a warm inspection room causes thermal expansion that changes the surface profile during the traverse. Allow 30-60 minutes of thermal stabilization before measuring to +/-0.001mm accuracy.
Insufficient traverse length. Measuring over fewer than five cutoff lengths violates ISO 4287 and reduces measurement reliability. On small features where the full evaluation length does not fit, reduce the cutoff length to the next lower tier and document the deviation.
Surface contamination. Coolant residue, chips, and fingerprints add false peaks to the roughness profile. Clean the surface with isopropyl alcohol and lint-free wipes before every measurement.

When supplier and customer readings disagree by more than 15%, mismatched cutoff length or filter type explains the discrepancy in most cases — lay direction error and stylus condition account for the rest. Verify these four settings before concluding the machining process is non-conforming.

Surface Finish by Manufacturing Process

Each process has a characteristic Ra band — finish turning typically lands at Ra 1.6 um while honing reaches Ra 0.2 um — so demanding Ra 0.4 um from a milling process forces a downstream grinding or honing step. Different processes produce characteristic roughness ranges. This table helps verify whether measured values are realistic for the process used and select the right process for the target specification. For guidance on optimizing cutting parameters to achieve specific Ra targets, see the CNC machining optimization guide.

Process	Achievable Ra Range (um)	Typical Production Ra (um)	Surface Character
Rough turning	6.3 - 25.0	12.5	Regular feed marks
Finish turning	0.4 - 6.3	1.6	Fine helical pattern
Face milling	0.8 - 6.3	3.2	Crossed arc pattern
End milling	0.4 - 6.3	1.6	Linear or helical
Cylindrical grinding	0.1 - 1.6	0.4	Fine random scratches
Surface grinding	0.1 - 1.6	0.8	Parallel scratch pattern
Honing	0.05 - 0.8	0.2	Crosshatch pattern
Lapping	0.01 - 0.4	0.05	Random, non-directional
EDM (fine)	0.8 - 6.3	3.2	Cratered, isotropic
Polishing	0.01 - 0.1	0.025	Mirror, non-directional

Ra Value (um)	ISO Grade	Common Applications
12.5	N10	Rough machining, non-critical surfaces
6.3	N9	General machined surfaces
3.2	N8	Standard machined finish
1.6	N7	Precision machined surfaces
0.8	N6	Bearing surfaces, seals
0.4	N5	Ground or lapped surfaces
0.2	N4	Precision grinding, honing

Honing and lapping are the only common processes capable of reaching Ra below 0.1 um in production — grinding typically bottoms out near Ra 0.1-0.2 um without additional superfinishing, and milling cannot reliably achieve Ra below 0.4 um even with sharp tooling and high feed rates.

For face-milling-side guidance on hitting a typical Ra 0.4-0.8 um with wiper inserts, see the face mill vs shell mill comparison.

When Ra Is Not Enough: Functional Surface Specification

Sealing surfaces need Rz, bearing surfaces need Rmr, and optical surfaces need waviness control — Ra alone serves none of these functions reliably. For critical applications, specifying Ra alone leaves too much to chance. Functional surface specification matches the parameter to the surface's job.

Sealing surfaces require Rz control. A hydraulic valve seat specified at Ra 0.4 um can still leak if Rz reaches 4.0 um from isolated deep scratches. Adding Rz max 2.5 um ensures no single valley creates a leak path. The ratio Rz/Ra for well-controlled processes is typically 4-7x; ratios above 8x indicate process instability.

Bearing surfaces require Rmr specification. The Abbott-Firestone curve defines the plateau, core, and valley zones of the profile. Specifying Rmr(c) = 70-90% at a cutting depth of 10% ensures adequate load-bearing area after running-in. Plateau honing of cylinder bores is the classic application -- the crosshatch valleys retain oil while the plateaued peaks carry the piston ring load.

Optical and decorative surfaces require waviness (W) control in addition to roughness. Waviness captures the longer-wavelength undulations that roughness filters remove. A surface can pass Ra 0.05 um yet show visible waviness that creates distortion in reflective applications.

Summary

Specify the parameter that matches the function, measure with the correct setup, and document both.

Ra is the starting point for surface finish control, not the complete answer. Add Rz for sealing surfaces where leak paths matter. Add Rmr for bearing surfaces where load-carrying area determines wear life. Select profilometer cutoff length from ISO 4288 based on the expected Ra range -- wrong cutoff is the leading cause of measurement disputes. Measure perpendicular to the lay direction, verify stylus condition, and allow thermal stabilization. When supplier and customer measurements disagree, compare instrument settings before questioning the machining process.

What is the difference between Ra and Rz surface finish parameters?

Ra is the arithmetic average roughness — it averages all profile deviations, masking isolated defects. Rz (ISO 4287) equals Rp + Rv: maximum peak height plus maximum valley depth within one sampling length, so a single deep scratch raises Rz without changing Ra much. Rz/Ra is 4-7× on well-controlled surfaces. For sealing surfaces, Rz is the more functionally critical parameter.

How do I select the correct profilometer cutoff length?

Per ISO 4288: for Ra 0.1-2.0 um, use 0.8mm cutoff with 4.0mm evaluation length. For Ra 2.0-10.0 um, use 2.5mm cutoff with 12.5mm evaluation length. The evaluation length always equals 5x the cutoff. Using the wrong cutoff shifts readings by 30-50%.

Why do my surface finish measurements differ from my customer's readings?

The three most common causes are: different cutoff lengths (check ISO 4288 compliance), different filter types (Gaussian vs 2RC produces 5-15% difference), and different measurement directions relative to the lay. Align these three settings and readings typically agree within 10%.

What does Rmr (material ratio) tell you about a machined surface?

Rmr describes the percentage of surface material above a specified depth. At a cutting depth of 10% of Rz, a high Rmr (70-90%) indicates a plateau-like surface with good load-bearing area, ideal for bearings and sliding contacts. A low Rmr indicates deep valleys with narrow peaks that wear quickly under load.

How do I specify surface finish correctly on an engineering drawing per ISO 1302?

Include the parameter and upper limit (e.g., Ra 1.6 um), the lay direction symbol if the pattern orientation matters, and the manufacturing process if required. For functional surfaces, add a lower limit to prevent over-finishing and specify Rz or Rmr in addition to Ra. Reference ISO 4288 defaults or call out specific cutoff and evaluation lengths.