Surface metrology

Surface roughness test methods and level development

A working guide to surface roughness test methods for engineers who have to accept or reject a machined part. We cover cut-off and filter settings, stylus and optical instruments, the parameters that matter on a drawing, and how the level system grew from early profile recorders to ISO 21920. Read it once and you can pick a method, set the instrument, and argue your case with a supplier.

Get Instant Quote Talk to an Engineer

Ra 0.2–3.2 μm±0.005 mmISO 4287 / 21920100% inspection

Surface roughness test methods applied to a CNC machined part surface

On this page

1
Key takeaways
2
What surface roughness test methods actually measure
3
Stylus, optical and comparison methods compared
4
Ra, Rz, Rq and Rmax, and when each one earns its place
5
How the roughness level system developed
6
Turning a measurement into an accept or reject decision
7
Step by step: run a surface roughness test on a machined part
8
Which surface roughness test method fits the job

Quick answer

Key takeaways

Match the method to the geometryStylus for flats and cylinders, optical for soft, thin or very fine surfaces.

Cut-off decides the numberThe same surface reports different Ra at 0.8 mm and 2.5 mm cut-off. Fix it on the drawing.

Ra alone hides peaksAdd Rz or Rmax when sealing, fatigue or coating adhesion is the concern.

Filtering is not optionalWithout a Gaussian or 2RC filter the waviness term inflates every roughness value.

State the standardISO 4287, ISO 21920 or ASME B46.1. Reports are not interchangeable.

Fundamentals

What surface roughness test methods actually measure

Every surface profile contains three scales at once: form, waviness and roughness. A surface roughness test method isolates the roughness band by filtering out the other two. The instrument does not measure roughness directly. It measures a profile with a tip or a light beam, then a filter strips the longer wavelengths away. Change the filter and you change the number on the screen.

That is why two labs can disagree on the same part and both be correct. A shop using a 0.8 mm cut-off on a turned shaft may report Ra 1.6 μm. A second shop using 2.5 mm cut-off on the same shaft may report Ra 2.4 μm. Neither lied. They answered different questions.

Before you buy an instrument, decide what the number is for. If it controls a bearing fit, you care about the plateau and the valleys. If it controls a paint or anodize bond, you care about the peak density and the open area between peaks. If it controls a seal, you care about the dominant wavelength of the tool marks. One parameter never covers all three.

1
RoughnessShort wavelengths from tool marks, grit and built-up edge, typically below 0.8 mm spacing.
2
WavinessLonger error from tool deflection, machine vibration and workholding.
3
FormThe intended geometry. Out-of-roundness and taper sit here, not in roughness.

Instruments

Stylus, optical and comparison methods compared

The contact stylus is still the reference method. A diamond tip, commonly 2 μm or 5 μm radius, is dragged across the surface at constant speed and the vertical movement is converted to a profile. It works on most metals, reads through coolant residue, and the trace is traceable to a calibration standard. The limits are real: a 2 μm tip cannot enter a slot narrower than about 2.5 μm, and it will scratch soft plastics and polished aluminium.

Optical methods split into three families. Focus variation and confocal microscopy scan a height map with no contact, so they suit soft polymers, thin-walled parts and fine finishes below Ra 0.1 μm. White light interferometry gives sub-nanometer vertical resolution but needs a reflective surface and a small field of view. Laser triangulation is fast and works inline, but its lateral resolution is coarse, so it is a trend tool rather than an accept-reject gauge.

Comparison methods still have a place on the shop floor. A tactile comparator plate or a visual standard lets an operator sort parts in seconds without a metrology lab. The trade-off is subjectivity: two operators will disagree near the boundary, and nobody can write a defensible report from a comparator. Use it for screening, then confirm anything borderline with a stylus trace.

1
Choose stylus whenThe drawing calls Ra or Rz, the part is metal, and you need a traceable report.
2
Choose optical whenThe material is soft, the feature is small, or the finish is below Ra 0.1 μm.
3
Avoid contact whenSurface damage is unacceptable, such as sealing faces on soft aluminium.

Parameters

Ra, Rz, Rq and Rmax, and when each one earns its place

Ra is the arithmetic mean deviation of the profile from the mean line over the evaluation length. It is stable, easy to compare and forgiving to a single scratch. Its weakness is that it averages everything. A surface with deep, sparse valleys and flat plateaus can carry the same Ra as a uniform ground surface, yet behave completely differently in a seal or a fatigue application.

Rz is the mean of the maximum peak-to-valley heights across the sampling lengths. It is far more sensitive to isolated defects. If a gasket has to crush into the valleys, Rz is the number that predicts leakage. Rq, the root mean square, weights peaks and valleys more heavily than Ra and correlates better with optical scatter and coating adhesion.

Rmax and Rt capture the single worst peak or valley in the evaluation length. Use them when one flaw can scrap the part: a hydraulic spool, a bearing race, a high-cycle fatigue fillet. The cost is repeatability. Rt changes every time you move the trace, so specify it only when you also specify the exact measurement position.

On a real drawing, a practical combination is Ra for the general finish, Rz for a sealing or coating face, and a note limiting the cut-off and evaluation length. One number with no cut-off is an incomplete callout, and it will be argued about at goods-in.

1
RaGeneral finish. Stable and comparable across suppliers.
2
RzSealing faces, crush joints and coated surfaces.
3
RqOptical parts, anodize, adhesion-critical surfaces.
4
Rmax / RtFatigue fillets and hydraulic fits, with a fixed trace position.

History

How the roughness level system developed

In the 1920s and 1930s, manufacturers began to notice that surface condition changed performance, and researchers in Germany, the United States and the United Kingdom built the first outline recorders. These machines drew a magnified trace of the surface on paper. Reading one was a skill, and comparing two traces was close to an argument.

The 1940s brought the first national standard, ASA B46.1 in the United States, which later became ANSI/ASME B46.1. It introduced the mean line system and a small set of parameters. National standards followed in Britain and Germany in the 1950s, all built on the same mean line idea but with different parameter names, which is why older drawings can be confusing.

From the 1960s onward, stylus instruments and optical methods made measurement routine, and the parameter set grew. ISO 4287 in 1982 consolidated profile parameters, and the 1990s added digital instruments and routine Ra and Rq reporting. Modern work has moved to ISO 21920 for profile specification, which reorganizes the older ISO 4287 and ISO 1302 callouts into one structure.

The practical lesson from that history is simple. The level system exists because a single number could not describe a surface. Each generation added parameters to close a gap that had already caused a field failure. When you write a callout today, you are choosing which of those gaps you care about.

Acceptance

Turning a measurement into an accept or reject decision

A measurement is not a verdict until it is compared against a stated rule. The rule has four parts: parameter, limit, cut-off and evaluation length. If any of the four is missing, the comparison is a judgment call. Write all four on the inspection plan and the argument at goods-in disappears.

Set the limit from function, not from habit. An Ra 1.6 μm callout on a non-functional cover adds cost with no benefit, while an Ra 0.4 μm callout on a hydraulic spool may still be too loose if Rz is uncontrolled. Ask what the surface has to do: seal, slide, bond, reflect or simply look uniform.

Handle borderline results with a repeat, not a renegotiation. Re-clean, re-level, run three fresh traces at a new position, and compare. If the part is genuinely at the limit, the range will stay wide and the peaks will stay high. If the first trace hit a burr, the second set will be tight and lower.

At GreatLight we inspect 100% of parts before shipment and run raw material checks, in-process monitoring and a final inspection, with reports on request. For finish-critical work we hold ±0.005 mm and can machine to Ra 0.2–0.8 μm, Ra 0.8–1.6 μm or leave an as-machined Ra 1.6–3.2 μm where the drawing allows it.

1
PassMean within limit and range narrow across three traces.
2
InvestigateMean near the limit or a single trace far above the others.
3
FailMean over the limit after a clean repeat at a new position.

Procedure

Step by step: run a surface roughness test on a machined part

1
Read the drawing firstWrite down the parameter, the numeric limit, the cut-off and the standard. If the callout says only Ra 1.6, ask which cut-off applies before you touch the part.
2
Clean and stabilize the partWipe with lint-free cloth and solvent, then let it reach room temperature. A part at 25 °C against a 20 °C instrument frame adds drift to a 2 μm tip trace.
3
Set the cut-off and evaluation lengthUse 0.8 mm cut-off for Ra 0.1–2 μm, 2.5 mm for Ra 2–10 μm, 8 mm above Ra 10 μm. Five sampling lengths is the default; three is acceptable on short features.
4
Level the part and orient the traceSet the traverse perpendicular to the dominant tool marks. A trace along the lay reads lower and does not represent the sealing face.
5
Pick the tip and the forceUse a 2 μm tip for fine finishes and a 5 μm tip for rough castings. Keep the measuring force low on aluminium and never use a 2 μm tip on a deep scratch.
6
Run three traces, not oneSpace them across the qualified area. Report the mean and the range. A single trace on a turned face can sit 30% away from the area average.
7
Check the filter and the outliersConfirm a Gaussian filter is active unless the drawing names 2RC. If one trace carries an obvious burr, note it rather than deleting it.
8
Record the setup with the resultLog instrument, tip radius, cut-off, evaluation length, filter, traverse direction and operator. A number without its setup cannot be repeated.

Selection

Which surface roughness test method fits the job

Match the method to material, finish level and the report you need.

Method	Best for	Typical range	Watch out for
Contact stylus 2 μm tip	Fine metal finishes, sealing faces	Ra 0.01–3 μm	Scratches soft aluminium
Contact stylus 5 μm tip	Rough castings, deep lay	Ra 0.5–20 μm	Misses narrow valleys
Focus variation	Soft polymers, thin walls	Ra 0.02–10 μm	Slow on steep slopes
White light interferometry	Polished and coated surfaces	Ra 0.005–1 μm	Needs reflective surface
Laser triangulation	Inline screening	Ra 0.5–50 μm	Coarse lateral resolution
Comparator plate	Shop-floor sorting	Ra 0.4–12 μm	Subjective at the boundary

Pick the method from the function, then lock the setup

Stylus with a stated cut-off covers most metal work. Go optical when the material is soft or the finish is below Ra 0.1 μm. Either way, write parameter, limit, cut-off and evaluation length on the drawing, because a number without its setup cannot be checked twice.

Get Instant Quote Talk to an Engineer

FAQs

Surface roughness testing questions we get

Can I compare an Ra value from a stylus with one from an optical scanner?

Only if both used the same cut-off, evaluation length and filter. Optical systems often report from an area rather than a line, which changes Ra by 10–30% on the same surface.

For a contract, name the instrument class and the parameters in the inspection plan. Otherwise you are comparing two different measurements that happen to share a symbol.

Why does my Ra reading change when I move the part?

The traverse is probably crossing a different lay direction, or the part is not level. A small tilt adds a slope that the filter only partly removes.

Re-level the part and rotate the traverse to sit perpendicular to the dominant tool marks. Repeat three times and compare the range.

Is Ra 0.8 μm always better than Ra 1.6 μm?

No. A smoother surface is not automatically a better one. Too smooth can starve an oil film, and it raises cost and cycle time.

Choose the finish from the function. If the surface has to hold a lubricant or grip an adhesive, a controlled rougher finish often outperforms a polished one.

How do I write a surface roughness callout on a drawing?

State the standard, the parameter, the limit, the cut-off and the evaluation length. One line is enough if it carries all five.

Example format: ISO 21920, Ra 1.6 μm, cut-off 0.8 mm, evaluation length 4 mm. Add Rz only where peaks and valleys drive function.

What is the difference between ISO 4287 and ISO 21920?

ISO 4287 defines the profile parameters themselves. ISO 21920 is the newer specification structure that covers how those parameters are called out and reported.

Older drawings reference ISO 4287 and ASME B46.1. New work increasingly references ISO 21920. Name one standard per drawing so the inspector knows which rule book applies.

Do you provide roughness measurement reports with machined parts?

Yes, on request. We record the parameter, instrument, tip radius, cut-off, evaluation length and filter alongside the result.

Parts are inspected 100% before shipment. If you need a specific trace position or a fixed evaluation length, put it on the drawing and we will measure to it.