Two round bars come off the same lathe, cut from the same heat of steel, to the same drawing dimensions. One runs to a fatigue life that lands neatly in the expected band. The other fails early, and the fracture starts at a shallow tool mark near the gauge length. Nothing on the dimension report separates them. The difference is the finish, and the number that usually governs it is Ra.
Surface roughness sits quietly on most specimen drawings, often in smaller print than the diameter and the radius. It still changes what the test measures, particularly for fatigue. This article covers what Ra is, why it moves tensile and fatigue results, how the requirement is written, and how we control it when we prepare specimens, so the finish on the part matches the finish the method assumes.
What Ra Actually Measures
Ra is the arithmetic average roughness of a surface profile. A stylus or optical sensor traces a short length of the surface, records how far the profile rises above and falls below a mean line, and averages those deviations. The result is a single number, usually in micrometres or microinches. A lower Ra means a smoother surface, and the general idea of surface roughness as a measurable, repeatable property is what lets a drawing turn a vague word like “smooth” into a target a shop can hit and an inspector can check.
Because Ra is an average, it hides as much as it shows. A surface with one deep isolated scratch can carry almost the same Ra as an evenly textured surface, even though the scratch is the more dangerous feature for fatigue. That is why drawings sometimes add a second parameter such as Rz or Rmax, which react to the peak-to-valley extremes rather than the average. When a callout lists more than one parameter, all of them are requirements, not alternatives. Reading Ra as the whole story is a common way to pass a number and still miss the intent.
Two other details travel with the number and are easy to overlook: the sampling length over which Ra is evaluated, and the lay, meaning the direction of the dominant tool marks. Both change how a surface behaves and how it measures. We come back to lay in a moment, because on fatigue specimens it can matter as much as the value itself.
Ra, Rz, and the Parameters That Travel With It
Ra earns its place because it is stable and repeatable, one number that does not swing wildly if the trace catches a slightly different spot. That stability is also its blind spot. Because it averages, it under-weights the isolated deep valley that a fatigue crack cares about most, and two very different surfaces can share an Ra.
That is why finish specifications for demanding parts often name a second parameter. Rz reports an average of the largest peak-to-valley heights across several sampling lengths, so it reacts to the extremes Ra smooths over. Rt captures the single largest peak-to-valley height in the evaluation length, the worst feature present. Rq, sometimes written RMS, is a root-mean-square version of Ra that weighs larger deviations more heavily and shows up on some older and some optical-instrument callouts. The broader vocabulary of surface finish parameters exists precisely because “average roughness” cannot describe every surface that matters. The practical takeaway is unchanged: when a drawing lists more than one, meet all of them, and do not quietly substitute the one your instrument reports by default.
Why Finish Changes Tensile and Fatigue Results
A rough surface is a field of small notches. Each valley raises the local stress above the nominal value, an effect described by stress concentration. Under a single monotonic pull, most ductile metals tolerate that, they yield locally and redistribute load, so a tensile test is comparatively forgiving of finish. You can still see finish influence where the fracture starts and, in brittle or high-strength materials, in the scatter of elongation and reduction-of-area numbers. For a routine tensile coupon, though, a reasonable machined finish is usually enough.
Fatigue is the opposite. Under repeated cycles, those same surface valleys become preferred sites for crack initiation, and fatigue cracks often begin at or near the surface. “Often” is the honest word here: subsurface initiation does occur, especially in high-strength alloys and very-high-cycle regimes, and inclusions or internal defects can win the race. But surface finish is one of the few crack-initiation variables the specimen preparer controls directly, which is why finish requirements tighten sharply when a part moves from static testing into fatigue. If you want the deeper treatment of fatigue specimen prep, we cover it in round fatigue specimen preparation for ASTM E466 and E606.
Finish also travels with what is underneath it. Aggressive machining or grinding can leave residual stress and a worked surface layer that shift fatigue behaviour independently of the roughness number. A part can read the right Ra and still carry a surface condition the drawing did not intend. That is a reason to control the whole finishing process, not just the final measurement, a point we return to below.
Where the Number Comes From
The finish requirement is not ours to invent, and it is not a default you can safely assume. It comes from the drawing, the material or product specification, and the test method being run. When those documents disagree, the drawing and the governing spec decide, not the shop and not the machine.
Fatigue methods are explicit that finish matters. ASTM E466, the practice for force-controlled constant-amplitude axial fatigue, calls for controlled specimen preparation and surface condition because the surface is where cycles do their early damage. ASTM E606, the strain-controlled fatigue method, is written around cyclic strain history that is controlled and measured, and it likewise depends on a well-prepared, consistent surface. Neither standard turns a roughness number into a guarantee of results; they set the conditions under which a valid comparison can be made. The official practices, ASTM E466 and ASTM E606, are the authority on the exact wording and current edition.
Static methods reference finish more lightly. ASTM E8/E8M, the general metallic tension method, and ASTM A370, which governs mechanical testing of steel products, are concerned with geometry, alignment, and gauge marking; they lean on the product specification and drawing for a specific surface callout rather than fixing one number themselves. So the rule of thumb is simple: read the drawing first, then the product spec, then the method. Do not carry a finish from a previous job onto a new part because the geometry looks similar.
How Smooth Is Smooth Enough
There is no universal Ra that makes a specimen correct, and any number quoted without its source should be treated with suspicion. The right value comes from the drawing and the governing method for that material and that test. What can be said in general is how the demand tends to scale with the job.
A routine machined tensile coupon is often accepted with a milled or turned finish in the range many general shops hold without a dedicated finishing pass. Move to fatigue, and callouts commonly tighten by an order of magnitude and add a lay requirement, because the surface has become a primary variable rather than a cosmetic one. Push into high-cycle fatigue on high-strength alloys, and the finish requirement can become one of the tightest features on the part. These are directional statements, not target numbers: your drawing sets the value, and a smoother-than-specified surface is not a free bonus, since it costs preparation time and can remove material you needed. The engineering aim is to match the finish to the requirement, then prove you matched it, rather than to polish reflexively toward a mirror.
Materials Where Finish Bites Hardest
Finish sensitivity is not uniform across materials. High-strength and hardened alloys tend to be the least forgiving, because their higher strength comes with lower tolerance for the local stress a surface valley creates, so a mark that a mild steel would shrug off can start a crack. Notch-sensitive materials behave the same way toward machining marks that act as small notches.
Welded specimens add their own wrinkle: the surface condition across the weld and heat-affected zone interacts with finish, and the drawing may govern both where the weld sits and how the surface is prepared. Specimens with directional microstructure, from rolling, drawing, or additive processes, can respond differently to a finish depending on how the lay lines up with that structure. None of this changes the workflow, but it does raise the stakes on getting the value, the lay, and the documentation right, and it is a good reason to confirm the requirement with the product owner rather than assume a familiar number carries over.
Reading a Surface-Finish Callout
A finish callout packs several requirements into one small symbol. It is worth slowing down on, because a specimen can meet the Ra number and still be wrong on the sampling length or the lay. Here is what each part of the symbol is telling you.
Read together, the four parts answer four different questions: is a finish controlled here at all, how smooth, measured over what length, and in which direction. Treat the value as a maximum unless the drawing says otherwise, and match your measurement cut-off to the one specified, or the same physical surface can report two different Ra values and you will chase a problem that is really a measurement setting.
Machining Marks, Lay, and Direction
On a round fatigue specimen, the difference between a good surface and a risky one is often direction, not depth. Marks left across the loading axis sit broadside to the stress and act as ready-made crack starters. Marks that run along the axis present far fewer transverse stress raisers to the same load.
This is why turned round specimens, which naturally carry circumferential (transverse) marks from the lathe, are frequently given a final longitudinal finish before fatigue testing. The goal is not only a lower number but a lay that runs with the load. We treat this as its own topic in longitudinal polishing for tensile specimens and why directional finish matters, because getting the direction right is often the higher-value move. A specimen polished to a slightly higher Ra with the correct lay can outperform a smoother one covered in transverse scratches.
How We Control Finish During Preparation
Controlling finish is a sequence, not a single machine. It starts with how the specimen is cut and continues through a dedicated finishing step, and each stage is where a number is either earned or lost. A blank extracted with excess heat or a torn cut surface can carry damage that later finishing has to remove rather than merely refine, so a clean extraction and a controlled machining pass make the finishing step shorter and more predictable. Trying to polish out a poor starting surface is slower, removes more material, and is harder to keep consistent from part to part.
For flat specimens, our flat sample preparation machines produce a controlled milled surface across the reduced section. Some configurations, such as the Classic Upgrade, are set up to hold surface roughness in the range many flat testing callouts ask for, subject to tooling condition, setup, and the chosen configuration. For round specimens, our round sample preparation systems turn the gauge and grip geometry; turning alone leaves circumferential marks, which is where the finishing step comes in. The trade-offs between the two families are covered in the difference between CNC machines for flat and round specimen preparation.
The longitudinal finishing step is what TensilePolish, our automatic longitudinal polisher system, is built for. It works the surface along the specimen axis under repeatable settings, which reduces the circumferential marks left by turning, controls the final lay, and standardises the finishing operation so it is not riding on one operator’s hand feel from part to part. What it does not do is remove the need to confirm the result: final Ra still depends on the material, the abrasive and settings chosen, and inspection of the finished part. We describe the outcome as a reduction of transverse marks and a controlled, documented finish, not as an automatic guarantee of a specific number or of residual-stress removal, which needs process-specific validation.
Measuring and Documenting Ra
Preparation and measurement are two different jobs. A specimen is not finished because the polisher stopped; it is finished when the surface has been measured where the drawing cares and the result recorded. Ra is typically read with a contact stylus profilometer or an optical instrument, and the two do not always agree to the last digit, so the method of measurement is part of the record.
Contact and optical instruments also see a surface differently. A stylus has a finite tip radius and can bridge the narrowest valleys, reading slightly smoother than an optical scan of the same spot; an optical system can be thrown by reflectivity on a bright finish. Neither is wrong, but a value only means something alongside the method that produced it, which is another reason the callout names a sampling length and the record names an instrument.
Measure where the load concentrates. On a fatigue or tensile specimen that is the reduced section or gauge length, not the grip ends, and taking several traces around and along that region is more honest than a single lucky pass. Match the sampling length to the callout, note the instrument and the lay direction, and keep the reading with the specimen identity. That record is what lets someone else trust the surface without re-inspecting it, and it is what turns “we polished it” into evidence.
Common Mistakes With Surface Finish
A few errors show up again and again, and most of them pass a casual check.
- Chasing a lower Ra than the spec asks for. A smoother surface costs prep time and can remove material; the target is the specified finish, not the smoothest one you can make.
- Meeting the number, ignoring the lay. A correct Ra with transverse marks on a fatigue part can still start a crack. Direction is part of the requirement.
- Measuring the wrong place or cut-off. A reading off the grip end, or at a sampling length that does not match the callout, describes a different surface than the one under test.
- Polishing that changes the part. Over-finishing can round a gauge edge or shift a diameter out of tolerance. Finish is a surface operation, not a place to remove size.
- An undocumented, inconsistent process. Manual finishing is not the problem by itself; a process that is not controlled or recorded from part to part is, because the next specimen may not match the last.
Matching Finish to the Test You Are Running
Finish requirements are not one-size-fits-all. The table below is a starting point for the questions to ask, not a substitute for the drawing and the governing standard, which always decide.
| Specimen / Test | Where Finish Usually Matters | Confirm Before Preparing |
|---|---|---|
| Static tension (E8/E8M, A370) | Moderate. A controlled machined finish is often adequate; watch fracture location on brittle or high-strength alloys. | Any explicit Ra on the drawing or product spec. |
| Axial fatigue (E466) | High. Surface is a primary crack-initiation site; value and lay both count. | Ra value, lay direction, and whether a final longitudinal finish is required. |
| Strain-controlled fatigue (E606/E606M) | High. Consistent, well-prepared surface supports valid cyclic strain data. | Finish callout plus gauge-section preparation requirements. |
| Round bar, turned only | Circumferential (transverse) marks from turning are the concern for cyclic loading. | Whether a longitudinal finishing step is specified before test. |
| Flat, milled reduced section | Milled finish across the gauge; edges and corners deserve attention too. | Ra on the reduced section and any edge-condition note. |
Getting a Repeatable Finish in Your Lab
The dependable path is a sequence you can repeat: cut and machine the geometry, apply a controlled finishing step in the required direction, measure Ra where the load concentrates, and record what you found with the specimen. Our flat and round preparation systems handle the machining, and TensilePolish handles the longitudinal finishing under repeatable settings, so the finish is a controlled step rather than an afterthought.
The number on the drawing is small, but it is doing real work: it is often the difference between a fatigue result you can defend and one that fails early for reasons no dimension report will explain. Treat Ra, its sampling length, and its lay as the requirement they are, and confirm the finished surface the same way you confirm a diameter. If you want help matching a preparation setup to the finish your specifications call for, that is exactly the conversation we are set up to have.