Surface Finish Chart: 2026 Conversion Guide for Manufacturers

North American drawings call out surface finish in microinches. European and Asian drawings use micrometers. Legacy prints use RMS or CLA. A single part crossing a global supply chain will hit at least two of those conventions before it ships.

This page is the reference for translating between them. It covers the conversion table, the arithmetic behind it, the roughness formulas, how to decode a full drawing callout, and which standards actually apply in 2026.

It does not tell you which finish to specify. For which machining process reaches which finish, and what each step down the scale costs, and for what a 125, 63 or 32 finish actually means in service, we have dedicated guides.

Surface Finish Conversion Chart: Ra, RMS, CLA, Rz, Rt and N Grades

This table covers every standard roughness grade from N1 to N12 in all six common notations, plus the cut-off length that governs how each band is measured.

How to read the chart

Find the row containing the one value you already have, then read across. A drawing calling out 125 µin is Ra 3.2 µm, 138.8 RMS µin, 125 CLA and N8, measured at a 2.5 mm cut-off.

The Ra columns and the CLA column are exact. The RMS column is an approximation valid for ordinary machined profiles. The Rz and Rt columns are indicative bands only, and the cut-off column is a requirement rather than a conversion.

Surface Finish Units of Measure

Five notations appear on real drawings. Two are current, two are legacy, one is a grade scale rather than a unit.

CLA deserves a note. It is not an approximation of Ra, it is Ra under an older name. If a 1970s British print says CLA 63, the modern spec is Ra 63 µin with no arithmetic applied.

Worked Conversion: a 125 Microinch Drawing and a Micrometer Inspection Report

This is the situation that generates the most queries, so here is the full arithmetic rather than a table lookup.

The drawing specifies Ra 125 µin maximum. The inspection report from the supplier reads Ra 3.4 µm. Does the part conform?

Step 1. Convert the drawing limit into the report's units. Multiply microinches by 0.0254:

125 µin × 0.0254 = 3.175 µm

Step 2. Compare like with like. The measured value is 3.4 µm. The limit is 3.175 µm. 3.4 is greater than 3.175, so the part is out of specification by 0.225 µm.

Step 3. Sanity check in the other direction. Convert the measured value back to microinches by multiplying by 39.37:

3.4 µm × 39.37 = 133.9 µin

133.9 against a 125 limit confirms the rejection. Both directions agree, which is the point of doing the check twice.

Note what nearly goes wrong here. The chart lists 125 µin against 3.2 µm, and 3.4 looks close to 3.2. Read casually against the tabulated row, this part passes. Read against the actual converted limit of 3.175 µm, it fails. Convert the limit, do not eyeball the row.

Surface Finish Conversion Calculator

Enter one value, pick its unit, and the equivalents calculate as you type. Rz is returned as a range because it is an estimate.

Surface Roughness Formulas, With Worked Numbers

Every parameter below is calculated from the same filtered roughness profile. They differ only in what they do with the height deviations. To make the relationships visible, all five examples use one eight-point profile with these deviations from the mean line, in micrometers:

+1.6, −1.5, +5.5, −1.6, +1.0, −1.0, +1.5, −5.5

Ra, arithmetic mean roughness

Ra = (1 / n) × Σ | y_i |

Sum the absolute deviations and divide by the count. Here: 1.6 + 1.5 + 5.5 + 1.6 + 1.0 + 1.0 + 1.5 + 5.5 = 19.2, divided by 8, so Ra = 2.40 µm.

Because Ra averages absolute values, one deep scratch barely moves it. Drop the −5.5 valley to −8.0 and Ra rises only to 2.71 µm, a 13 percent change from a defect that more than doubles the depth of the deepest feature on the surface. That is why Ra can pass a surface that a seal will still leak across.

Rq, also written RMS

Rq = √ [ (1 / n) × Σ y_i² ]

Squares: 2.56, 2.25, 30.25, 2.56, 1.0, 1.0, 2.25, 30.25. Sum 72.12, divided by 8 gives 9.015, square root gives Rq = 3.00 µm.

Divide that by the Ra of 2.40 and you get 1.25, not the 1.11 that the conversion charts use. Both figures are correct, and the gap between them is instructive rather than an error.

Rq squares the deviations before averaging, so it weights the extremes more heavily than Ra does. How much more heavily depends entirely on how many points sit near the extremes. An eight-point illustration is far too coarse for two large spikes to be diluted, so the ratio runs high. A real profilometer trace carries thousands of sampled points across the evaluation length, the extremes make up a tiny fraction of them, and the ratio settles near 1.11 for ordinary machined surfaces. The 1.11 factor is therefore a statistical property of dense real profiles, not a defined constant and not something a small worked example will reproduce. It drifts on any surface with an unusual peak distribution, which is the reason to measure rather than convert when the spec matters.

Rt, total height of the profile

Rt = Zp + Zv

The highest peak plus the deepest valley across the whole evaluation length. Here the highest peak is +5.5 and the deepest valley is −5.5, so Rt = 11.0 µm. Rt is driven entirely by the two worst points on the surface.

Against an Ra of 2.40 that is a ratio of 4.58, which sits inside the Ra × 4 to Ra × 10 band in the chart above. A profile with the same Ra but sharper isolated spikes would push that ratio toward the top of the band, which is precisely why the band is wide and why Rt is not converted from Ra.

Rz, mean roughness depth

Rz = (1 / 5) × Σ Rz_i

The evaluation length is split into five sampling lengths. Each contributes its own peak-to-valley height, and Rz averages those five. For this surface the five come out at 9.8, 11.0, 9.2, 10.4 and 9.6 µm. Sum 50.0, divided by 5, gives Rz = 10.0 µm.

Two checks are worth making on those numbers. First, Rz of 10.0 µm is below Rt of 11.0 µm, as it must be: Rt takes the single worst peak and the single worst valley anywhere in the evaluation length, so it can never come out lower than an average of five section heights taken from that same length. Second, the second sampling length returns 11.0 µm, equal to Rt, because that is the section in which both the +5.5 peak and the −5.5 valley happen to fall.

Against an Ra of 2.40, Rz of 10.0 is a ratio of 4.17. On a surface with more pronounced peaks the same Ra could carry an Rz of 15 or more. This is why Rz is never converted from Ra.

Rmax, also written Rz1max

Rmax = the largest single Rz_i

From the same five values, the largest is Rmax = 11.0 µm. Rmax is the worst sampling length rather than the average of all five, which makes it the parameter of choice for sealing faces where one bad section fails the part. On this profile Rmax and Rt coincide at 11.0 µm because the two extremes share a section. When the highest peak and deepest valley fall in different sections, Rmax comes out below Rt, and the ordering Ra ≤ Rq ≤ Rz ≤ Rmax ≤ Rt always holds.

Theoretical turning finish

Rz (µm) = ( f² / (8 × r) ) × 1000

With feed f in mm/rev and nose radius r in mm. At f = 0.2 and r = 0.8: 0.04 / 6.4 = 0.00625 mm, so Rz = 6.25 µm.

This is a floor, not a forecast. It models the geometry of the feed marks and ignores tool wear, built-up edge, vibration and workpiece rigidity. Measured values are always worse, and the gap widens as the insert dulls. Use it to choose a starting feed, then verify on the part.

How to Read a Full Surface Texture Callout

Most callouts carry a single Ra number. A fully populated one carries six fields around the checkmark, and each field is a separate requirement that must be verified on its own.

The two standards place the fields in broadly the same positions, but they do not mean the same thing in every slot. The difference that catches people out is the field to the right of the roughness value. Under ASME Y14.36 that area conventionally carries waviness, height and spacing. Under ISO 21920-1 it carries a second roughness parameter. A great many online symbol diagrams show the ISO meaning on an ASME-labelled figure, which is why the same callout gets read two different ways by two different shops.

Decoding an ASME callout, field by field

Take a callout with a bar across the top and these entries: a = 1.6, b = GRIND, c = 0.8, d = ⊥, e = 2, f = 0.05-2.

The bar makes material removal mandatory, so the surface cannot be left as cast. Position a sets Ra 1.6 µm as the upper limit. Position b names grinding as the required process, which removes the supplier's freedom to reach 1.6 µm by fine turning instead. Position c fixes the cut-off at 0.8 mm, and that matters, because measuring the same surface at 2.5 mm returns a different number. Position d requires the lay to run perpendicular to the edge shown in that view. Position e reserves 2 mm of stock for the grinding operation.

Position f is the one most readers skip. Under ASME it reads as waviness height 0.05 mm and waviness spacing 2 mm. That is a separate control on longer-wavelength form error, and a roughness measurement will not detect a violation of it. A ground surface can sit comfortably inside Ra 1.6 µm while still failing waviness because of chatter or wheel deflection, and a profilometer set to report Ra alone will pass the part.

The same field under ISO

On an ISO 21920-1 drawing that slot means something different. An entry of Rz 6.3 in position f adds an independent roughness limit alongside the Ra limit in position a. Ra 1.6 with Rz 6.3 is a ratio just under 4, which is tight. Both must be met, and neither implies the other.

ISO 21920-1 also allows far more than the older ISO 1302 syntax could express, including explicit filter types and nesting indices, evaluation length, section count, and which tolerance acceptance rule applies, the maximum rule, the 16 percent rule or the median rule. A complete ISO 21920 indication can therefore be considerably longer than the ASME equivalent. If you receive a callout carrying a U or L prefix, or a filter designation, it is written to ISO 21920 and should not be read using ASME field conventions.

Two modifiers change the meaning of the whole symbol under both standards. A circle at the base forbids material removal, so the surface stays as cast, forged or moulded. A bar across the top requires it. A plain checkmark with neither leaves the method to the shop. For the checkmark geometry and the full lay-direction table, see our guide to reading surface finish symbols on a drawing.

Four Conversion Mistakes That Scrap Parts

These are the errors our quality team actually catches on incoming drawings and supplier reports, in order of how often they appear.

Treating RMS as Ra

The 1.11 factor looks small enough to ignore until it decides a rejection. A drawing calling out 63 RMS is asking for Ra 56.8 µin, not Ra 63. A supplier who machines to Ra 63 and reports it as compliant is delivering roughly 70 RMS, which is 11 percent over the limit. On a hydraulic sealing face that is the difference between a part that seals and a warranty claim. Convert first, compare second.

Assuming Rz or Rt is a fixed multiple of Ra

The 4 to 7 band for Rz and the 4 to 10 band for Rt are observations across machined surfaces, not rules. Two parts with identical Ra can differ by a factor of two in Rz, because Ra averages the whole profile while Rz and Rt key off the extremes. A lapped surface sits near the bottom of both bands, a milled surface with intermittent chatter near the top or above it. The one relationship you can rely on is the ordering: Rt is never less than Rz on the same evaluation length. If a drawing specifies more than one parameter, measure each one.

Transposing microinches and micrometers

The costliest and simplest error. Ra 125 µin is a routine machined finish. Ra 125 µm is 39 times rougher and is not a producible machining spec at all. The tell is the decimal point: metric callouts carry one (1.6, 3.2, 6.3) and imperial callouts generally do not (63, 125, 250). A bare "125" on an unlabelled drawing should be queried, not assumed.

Comparing values measured at different cut-offs

Cut-off length is not metadata, it is part of the specification. Measuring a surface at 0.8 mm when the drawing calls for 2.5 mm filters out longer-wavelength content and returns a lower Ra on a physically unchanged part. If a supplier report and an incoming inspection disagree by 20 or 30 percent, check the cut-off on both before investigating the process.

How Surface Roughness Is Measured

Four methods, chosen on accuracy needed, whether the surface can be touched, and whether the check happens at the machine or in the lab.

Contact profilometer. A diamond stylus traverses the surface and records the profile, from which the instrument computes Ra, Rq, Rz, Rt and Rmax. This is the reference method and the one ASME B46.1 and ISO 21920 are written around. Limitations: it samples a single line rather than an area, and the stylus can mark soft or coated surfaces.

Non-contact optical. Confocal microscopy, white light interferometry and laser triangulation map an area in three dimensions without touching the part. Standard for soft metals, coatings, and any finish below Ra 0.1 µm where stylus radius becomes a limiting factor.

Comparator blocks. Reference plates at known Ra, compared by sight and fingernail. Fast and genuinely useful for a first-pass sort on the floor. Not evidence, and never acceptable on an inspection report.

Portable testers. Handheld stylus units taken to the machine for in-process checks between setups.

Sampling length and cut-off

The cut-off wavelength λc separates roughness from waviness. Set it too short and real roughness is filtered out as waviness. Set it too long and form error contaminates the reading. ISO 4288, withdrawn and replaced by ISO 21920-3, ties λc to the expected Ra band: 0.08 mm for Ra at or below 0.02 µm, 0.25 mm from Ra 0.02 to 0.1, 0.8 mm from Ra 0.1 to 2, 2.5 mm from Ra 2 to 10, and 8 mm above that. The chart above lists the cut-off for each standard grade from N1 up. The standard evaluation length is five consecutive sampling lengths.

Where a drawing is silent, the default is Ra measured over five sampling lengths at a 0.8 mm cut-off. If your part sits outside the Ra 0.1 to 2 band, that default is wrong for your surface and should be stated explicitly on the print. Our team verifies finish on a calibrated profilometer as part of first article inspection on every custom CNC machining programme, and records the cut-off used on the report so two labs can reproduce the same number.

Surface Finish Standards in 2026

Four documents matter, and one of the four is still cited on new drawings years after its withdrawal.

What is ASME Y14.36M?

ASME Y14.36 is the North American standard for surface texture symbols. It defines the checkmark, the field positions, the lay symbols, and how roughness and waviness are annotated on a drawing. It deliberately does not specify how the surface is produced or how it is measured, which is B46.1's job.

The designation most people search for, Y14.36M, is the 1996 metric-designated edition. It remained in effect for well over a decade and is still quoted on a large number of legacy prints. The current edition is ASME Y14.36-2018, now under stabilized maintenance, meaning ASME will consider change requests but is not actively revising it. The M was dropped from the designation. If a drawing in front of you cites Y14.36M-1996, it is not wrong, it is old, and the symbol conventions you need are substantially unchanged.

What is ASME B46.1, and is ANSI B46.1 the same thing?

ASME B46.1 is the companion measurement standard, currently the 2019 edition. Y14.36 says how to draw the requirement. B46.1 says how to measure it, defining Ra, Rq, Rz, Rt, the waviness parameters, instrument setup and filtering.

ANSI B46.1 and ASME B46.1 are the same document. It is an ASME standard accredited by ANSI, so both prefixes circulate. Older references and many supplier quality manuals use ANSI, current ones use ASME. There is no separate ANSI version to obtain.

ISO 1302 is withdrawn: what replaced it

This is the correction most surface finish reference pages have not made. ISO 1302:2002 was withdrawn and superseded by ISO 21920-1:2021. The same revision withdrew ISO 4287 and ISO 4288, replacing them with ISO 21920-2 for terms and parameters and ISO 21920-3 for specification operators.

Practically, drawings issued before 2022 will keep citing ISO 1302 and ISO 4287 for years, and those callouts remain readable. Ra, Rz and Rq survive into ISO 21920-2 with their definitions substantially intact, so the numbers on an old print still mean what they meant. What changed is the specification syntax and the default rules around sampling and filtering. ISO 21920-1 introduces explicit tolerance acceptance rules, the maximum rule, the 16 percent rule and the median rule, along with filter types and nesting indices that ISO 1302 had no way to express. A drawing written to ISO 21920 can therefore state requirements an ISO 1302 drawing could only leave to default.

If you receive a print citing ISO 1302, read it as written. If you are issuing a new print, cite ISO 21920-1.

When a programme runs across our facilities for a North American customer, inspection reports map both systems so the ASME and ISO numbers on the same part agree rather than being argued about at receiving.