Key takeaways
- Ra is the arithmetic mean of the profile ordinates and Rz is the sum of the highest peak and deepest valley in one sampling length, so on a sine profile the ratio between them is exactly π, or 3.14.
- A 20 µm deep, 50 µm wide scratch across a 2.5 mm sampling length raises Ra by only 0.2 µm while taking Rz to 20 µm, which is why Ra alone cannot police a sealing face.
- An Ra number without a cut-off is not reproducible: 2.5 mm is the common default, measured over five sampling lengths, giving a 12.5 mm evaluation length.
- A typical fibre-laser cut edge on 3 mm mild steel sits around Ra 2–6 µm and Rz 10–25 µm depending on assist gas and speed, and no drawing should assume a polished finish from a cut edge.
- Call Rz where the surface seals, bonds or coats, call Ra where it is a general finish requirement, and call a visual standard where the requirement is cosmetic.
A surface roughness measurement is only meaningful when the drawing says which parameter, over what length, in which direction. Ra on its own is an average, and averages are good at hiding the single deep striation that lets a gasket weep or a coating blister.
The arithmetic is blunt. A 20 µm deep, 50 µm wide scratch crossing a 2.5 mm sampling length adds about 0.2 µm to Ra, and takes Rz to 20 µm. A drawing that calls Ra 3.2 µm accepts that part. A drawing that calls Rz 12 µm rejects it. Both callouts describe the same surface, and only one of them describes the failure the buyer cares about.
What follows sets out what Ra and Rz each measure, the ratio between them and why it moves, the cut-off that makes a number reproducible, and the calls that suit a laser-cut edge.
What Ra and Rz actually measure
ISO 21920-2:2021 sets the terms and parameters for profile surface texture, and it replaced ISO 4287, which is withdrawn. Drawings still carry "Ra per ISO 4287", and that is acceptable as long as the edition is stated, but a new drawing should reference the current standard so that the definitions are unambiguous.
Ra is the arithmetic mean of the absolute ordinates of the roughness profile over the sampling length, measured from the mean line. Every point on the trace contributes, weighted by how far it sits from the line. A single deep valley is one point among thousands, so it moves Ra by very little.
Rz is the maximum height of the profile within a sampling length: the height of the highest peak plus the depth of the deepest valley. It is not an average, and it is dominated by exactly the feature that Ra ignores. Over a full evaluation length, the largest of those values is often reported separately as Rmax or Rt.
That difference is the whole argument. A general finish call is a statement about the typical surface, and Ra describes it well. A sealing, bonding or coating call is a statement about the worst feature, and only Rz or Rmax describes it.
The ratio between Ra and Rz is not a constant
Plenty of drawings and inspection reports treat Rz as roughly four times Ra, and use that to convert. The ratio holds for some surfaces and fails for others, and the failure is where the arguments come from.
For a sine profile of amplitude A, the arithmetic mean of the absolute ordinate is 2A/π, which is 0.637A, and Rz is 2A. The ratio is therefore π, or 3.14. For a turned or milled surface with a regular tool mark, the ratio usually lands between 4 and 6. For a laser-cut edge where striations are irregular and there is occasional dross, the ratio can reach 6 to 8, because a few deep valleys carry the Rz while the rest of the profile is comparatively smooth.
The practical consequence is that converting one to the other is guesswork. If a drawing calls Ra 3.2 µm and the process produces a regular profile, Rz lands near 13 µm. If the same Ra comes from an irregular edge with one deep striation period, Rz can be 25 µm. Same Ra, twice the peak-to-valley, and a different answer at the gasket.
What a cut-off does to the number
A roughness instrument separates waviness from roughness with a cut-off, and the cut-off decides what counts as roughness. The common default for non-periodic profiles is a 2.5 mm cut-off, with the evaluation length set at five sampling lengths, so 12.5 mm of trace. A finer 0.8 mm cut-off is used where the surface is fine, and an 8 mm cut-off where the process leaves long waves.
Change the cut-off and the number changes. A 1 µm amplitude waviness at a 5 mm wavelength is filtered out at a 2.5 mm cut-off and included at 8 mm. Its contribution to Ra, if included, is 0.637 × 1 = 0.64 µm, which is enough to take a surface reading 0.8 µm to 1.4 µm without anything on the part changing.
This is why a call of "Ra 1.6" with no cut-off is not a specification. The supplier measures to the instrument default, the customer measures to their own procedure, and both report honest numbers that disagree. State the parameter, the value, the cut-off, the evaluation length and the direction, and the argument disappears.
Where each parameter belongs
Ra suits a general finish requirement on a stable process: a visible panel that has to look even, a bracket face that has to sit flat against a mating part, a surface that will be painted and where the paint, not the steel, is the finish. It is easy to measure, robust to a single bad point, and it correlates with how a surface looks and feels.
Rz suits anything where the extremes do the work. A gasket needs a surface whose peaks do not have to be crushed flat before the seal closes, and a deep valley becomes a leak path. An adhesive bond needs a surface that the adhesive can key into, and a peak that stands above the bond line is a void. A coating needs a profile that the film can cover, and a valley deeper than the film is a spot where the steel is bare.
Rmax, or Rt over the evaluation length, suits the cases where one feature anywhere on the surface is enough to fail: a sealing groove, a bearing face, a hydraulic port. It is the most sensitive to a single scratch, and the least stable between measurements, which is why it is used as a limit rather than as a process control.
Edge finish on a laser-cut part
| Parameter | What it measures | Typical on a 3 mm mild steel laser edge | Ganesh limit or note |
|---|---|---|---|
| Ra | Mean deviation of the profile over the sampling length | 2–6 µm, by gas and speed | Achieved as cut on the 12 kW fibre laser; below 1.6 µm needs a secondary operation |
| Rz | Highest peak plus deepest valley in one sampling length | 10–25 µm | The number to call on a sealing or coating face |
| Rmax / Rt | Largest peak-to-valley over the evaluation length | 15–35 µm | Used as a limit where one striation can fail the part |
| RSm | Mean width of the profile elements | 0.1–0.4 mm on a striated edge | Describes striation spacing, which drives gasket contact |
| Cut-off λc | Filter separating roughness from waviness | 2.5 mm default, 0.8 mm for fine faces | Must be stated on the drawing, with the evaluation length |
| Evaluation length | Trace length used for the reported value | 12.5 mm at five sampling lengths | Shorten it and Rmax falls; the limit has to fix it |
| Direction | Trace across or along the striations | Across gives the higher Rz | State it, because the two directions disagree on the same edge |
| Post-treatment | Deburr, blast or brush | Blasting raises Rz by 10–20 µm | A finish call made before blasting will not survive it |
| Coating build | Film over the peaks after powder | 60–120 µm typical | Read with Rz: a film thinner than Rz leaves peaks exposed |
The table makes one point that buyers tend to resist: a laser-cut edge is a cut edge, not a machined surface. The striations are a record of the beam moving through the material, and their depth depends on assist gas, focus position, power and speed. Asking for a lower Ra on the same edge is asking for a different cut, and past a point it is asking for a different process.
Roughness, blasting and coating
Powder coating does not smooth a surface, it follows it. A 60 µm film over a surface with Rz 25 µm covers the profile with 35 µm to spare at the valleys and a thinner layer over the peaks. A 60 µm film over Rz 40 µm leaves the peaks close to the surface, and that is where a coating fails first in a salt spray or a humidity test.
This is also why a DFT reading on a blasted surface has to be corrected. ISO 19840 deals with measuring dry film thickness on rough surfaces, and its whole premise is that the gauge reading includes the profile below the film: the thickness that protects the steel is the reading above the peaks, not above the valleys. A plant that reads 90 µm on a blasted panel without the correction can be over-reporting by 10–20 µm depending on the profile.
The order of operations matters as much as the numbers. Blasting in the 16 ft × 16 ft chamber raises Rz before the seven-tank pretreatment, pretreatment leaves its own micro-texture, and the powder goes on last. A drawing that calls Ra on the incoming steel and Rz on the finished part is asking for two measurements of two different surfaces, and both have to be stated.
What Ganesh runs
Edges are cut on the 12 kW fibre laser with its 40 ft × 16 ft bed, with 6 kW and 3 kW support machines, in mild steel, CRCA, stainless, aluminium, brass and copper up to 20 mm plate. Thickness range for custom parts runs from 0.8 mm to 40 mm, and the edge a drawing sees is a function of the gas, the focus and the speed set for that thickness and material.
The plant does not polish, grind or machine surfaces as a standard operation. Where a drawing calls a finish that a cut edge cannot produce, the route is a secondary operation quoted separately, and it is better to know that at quotation than at first article. Deburring is standard on parts that handle, and it changes the edge profile at the corners without changing the face roughness.
Coating runs through the seven-tank pretreatment line, the 16 ft × 16 ft blast chamber and the 18 ft × 14 ft powder booth, with film build checked by DFT gauge and adhesion by cross-hatch. The system is certified to ISO 9001:2015 and ISO 14001:2015, and the same records carry the roughness call and the coating call where both appear on the drawing.
What to put on the drawing
- The parameter: Ra, Rz or Rmax, with the value in µm and the unit written out.
- The standard and edition the parameter is defined by, so that the definitions do not drift.
- The cut-off and the evaluation length, or a reference to the standard's default procedure.
- The location: which face or edge, and the distance from the cut edge where the trace starts.
- The direction of measurement, relative to the striations or the rolling direction.
- Whether the value is a maximum or a nominal, because a maximum on Rmax is a different requirement from a typical value on Ra.
- The post-treatment, since blasting or brushing changes the number after the cut.
- The coating build, where the surface is coated, so that the roughness and the film are read together.
When Ra is the wrong call
Ra is the wrong call on a sealing face, because the peak that breaks the seal contributes almost nothing to the average. It is the wrong call on a bonded joint, where the adhesive needs a profile it can key into and the average says nothing about whether that profile exists. It is the wrong call on a cosmetic class A panel, where the requirement is what the eye sees under a light at a stated angle, and no profile parameter captures a dent or a smudge.
It is also the wrong call where the surface has been blasted. The blast profile is described by Rz and by peak count, and an Ra call on a blasted surface is a measurement of the average of a texture that was chosen for its peaks. The same applies to a laser-cut edge that will be powder coated: the number that predicts the coating performance is Rz, not Ra.
There is a commercial version of the same mistake. A drawing that calls Ra 0.8 µm on every edge of a laser-cut bracket has specified a polished part and priced a cut one, and the difference shows up as a rejection at first article. Call the finish where the function needs it, and leave the rest as cut. Send the drawing with the edge and finish callouts, and the engineering team will confirm what the cutting and coating route can hold before the price is fixed. Request a quote.
The plant and quality hub carries the measurement side of the system. The laser-cut edge quality guide covers how thickness and gas change the edge, the deburring guide covers what happens after the cut, and the coating thickness guide takes the roughness number through to the film. Laser-cut parts are quoted through the laser cutting capability.
Frequently asked questions
Should a drawing call Ra or Rz?
Call Ra for a general finish requirement on a stable process and Rz where the surface has to seal, bond or take a coating. Rz captures the isolated peak or valley that Ra averages away, and a 20 µm scratch barely moves Ra.
What cut-off should be specified with Ra?
State the cut-off and the evaluation length, for example Ra 3.2 µm with a 2.5 mm cut-off over a 12.5 mm evaluation length. Without them, two laboratories can report different values for the same surface and both be correct to their own procedure.
What Ra can a laser-cut edge hold?
A fibre-laser cut edge on 3 mm mild steel typically falls between Ra 2 µm and 6 µm, with Rz between 10 µm and 25 µm, depending on assist gas, focus and speed. A call below Ra 1.6 µm on a cut edge needs a secondary operation, not a different cutting parameter.
Does powder coating hide a rough edge?
It covers it, but the peaks still matter. A 60 µm film over a surface with Rz 40 µm leaves thin coverage at the peaks, which is where corrosion starts. The roughness call and the film build have to be read together.
Why do two inspectors get different Ra on the same part?
Usually a different cut-off, a different evaluation length, a different measurement direction, or a stylus that has picked up a burr at the edge. Agree the procedure on the drawing and measure at a stated distance from the cut edge.
Is Ra the same as RMS?
No. Ra is the arithmetic mean of the absolute ordinates and RMS is their root mean square, which weights the peaks more heavily. On a sine profile RMS is about 11% higher than Ra, so the two are close but not interchangeable.
Sources and further reading
- ISO 21920-2:2021 — Geometrical product specifications, surface texture: profile parameters · reviewed 3 Nov 2026
- ISO 19840:2012 — Measurement of, and acceptance criteria for, dry film thickness on rough surfaces · reviewed 3 Nov 2026
- CSIR-NPL — National Physical Laboratory, India · reviewed 3 Nov 2026
Send the edge and finish requirement
State the parameter, the cut-off and where the measurement is taken. The engineering team confirms what the cutting route can hold before quotation.
Request a quote