Surface Texture & Roughness — Engineering Decision Guide

Only tighten finish where a function demands it, and set the target from that function: roughly Ra 0.1–0.4 µm for dynamic seals, sliding/bearing surfaces, and fatigue-critical parts; Ra 0.4–1.6 µm for general precision fits and static seals; Ra 1.6–3.2 µm for typical machined surfaces; and Ra 3.2–6.3 µm+ where finish is cosmetic or non-critical. Leave non-functional surfaces coarse.

Surface finish drives friction, wear, sealing, fatigue life, and coating adhesion — but each finer band costs more machining, grinding, or polishing time. A dynamic seal needs a smooth, plateaued surface so the lip doesn't leak or wear; a fatigue-critical fillet needs smoothness because roughness valleys are stress raisers that start cracks. Conversely, a surface that only needs to look acceptable gains nothing from a mirror finish. Specifying to function is what keeps cost proportional to need.

Too rough on a seal or bearing leaks and wears fast; too rough on a fatigue fillet cracks. Too fine everywhere multiplies cost for no benefit — over-specifying finish is a classic silent cost driver. A number with no function behind it invites both.

Use Ra (average roughness) as the general default for process control and most callouts. Add or switch to Rz (mean peak-to-valley) when peaks and valleys matter — sealing, coating, fatigue — because it catches extremes Ra averages away. Use Rt/Rmax (maximum height) where a single flaw is unacceptable (a deep scratch on a seal land). For functional surfaces, consider bearing-ratio (Rk-family) parameters.

Ra is an average over the profile, so it's stable and repeatable — but two very different surfaces can share the same Ra: one gently wavy, one with sharp deep scratches. Rz and Rt capture the peak-to-valley extremes that Ra hides, and those extremes are exactly what leak past a seal, concentrate stress, or hold a coating. That's why a seal or fatigue surface is often specified with Rz (or a max limit), not Ra alone.

Specifying Ra alone on a sealing or fatigue surface can pass a part that has damaging deep scratches (same Ra, worse Rz). Using Rt for routine process control makes it noisy and hard to hold. The parameter has to match what the function actually cares about.

Pick the process that reaches the target economically: milling/turning for ~Ra 0.8–6.3 µm, grinding for ~Ra 0.1–0.8 µm, honing for cylinder bores (~Ra 0.1–0.4 with a crosshatch), and lapping/superfinishing/polishing for <Ra 0.1 µm mirror finishes. If the target is finer than the primary process reaches, plan a secondary finishing step rather than chasing it with the wrong tool.

Each process has a practical finish range set by how it removes material — a turned surface carries the feed-mark scallops, grinding leaves finer abrasive scratches, lapping/polishing removes almost nothing but the peaks. Trying to reach a grinding finish by turning means impractically light feeds and still may not get there; the economical path is to rough with one process and finish with another. Honing's crosshatch is specifically engineered to hold oil in cylinder walls, which a smoother finish wouldn't.

Demanding a mirror finish from a milling operation balloons cycle time and often still misses. Specifying a finish no economical process reaches drives cost or scrap. Removing the honing crosshatch (too smooth a bore) can starve lubrication.

Specify the cutoff (sampling length) along with the Ra value — the standard default scales with roughness (e.g. 0.8 mm cutoff for typical machined finishes), and it must match between drawing and measurement. Where it matters, call out the lay (the direction of the dominant machining pattern) and measure across the lay, since that's where roughness is greatest and most functionally relevant.

The cutoff is the wavelength filter that separates roughness from waviness — a longer cutoff includes more of the longer-wavelength variation and reports a higher number, so the same surface measured at different cutoffs gives different Ra. That's a leading cause of supplier/customer measurement disputes. Lay matters because roughness measured along the tool marks reads far smoother than across them; sealing and sliding functions care about the across-lay value.

An unspecified cutoff lets two labs measure the same surface and disagree. Measuring along the lay instead of across understates roughness and can pass a functionally rough surface. Ignoring lay on a directional sliding/sealing surface misses the parameter that governs performance.

The same surface reads differently on two instruments or between supplier and customer, or a part fails finish inspection that looks and functions fine. Arguments over whether a surface passes, with both sides "right" on their own setup.

Different cutoff/sampling length between measurements, measuring along vs across the lay, a different parameter (Ra vs Rz), stylus tip radius/condition differences, and instrument calibration or a dirty/nicked surface. Cutoff and direction are the usual culprits.

1) Fix and share the full spec — parameter, value, cutoff, direction — so both sides measure the same way. 2) Measure across the lay unless the drawing says otherwise. 3) Calibrate the instrument on a reference standard; check stylus condition. 4) Clean the surface and avoid measuring over a nick or handling mark. 5) Agree the method up front for critical surfaces to head off receiving-inspection disputes.

A seal weeps, a sliding surface galls or wears fast, a coating peels, or a highly loaded part cracks early from the surface — despite passing its dimensional checks. The size was right; the texture was wrong.

Too rough for a dynamic seal (leak path and lip wear), too smooth or wrong texture for lubrication (galling from no oil retention), roughness valleys acting as fatigue crack initiators on loaded parts, and poor surface prep hurting coating adhesion.

1) Set the finish (and parameter) from the function — seal, slide, fatigue, or coat — not a generic default. 2) For seals, control both roughness and peak/valley (Rz) and often the lay/plateau. 3) For lubricated sliding, keep enough texture to retain oil (e.g. honing crosshatch). 4) For fatigue parts, specify a smooth surface in the critical fillet/zone. 5) For coatings, match the prep/roughness the coating requires for adhesion.