Surface finish chart: Ra, RMS and N grades

No — they are two different averages of the same profile, and one cannot be converted into the other exactly. Ra is the arithmetic mean of the deviations from the mean line; RMS, properly Rq, is the root mean square of them. For a perfect sinusoid the ratio is exactly π ÷ (2√2) = 1.1107, which is where the familiar eleven percent comes from, but a machined surface is not a sinusoid: squaring the deviations weights the outliers, so a ground finish with occasional deep scratches lands well above the ratio while a regular turned profile sits near it. Where a drawing calls out RMS, measure Rq rather than converting, and be aware that older US prints marked RMS often meant Ra because the instruments of the period computed Ra.

Because the microinch column of the grade series is a conventional rounded figure, not an arithmetic conversion. The N grades were defined in micrometers as a doubling series rounded to preferred numbers — 0.4, 0.8, 1.6, 3.2, then 6.3 rather than 6.4 and 12.5 rather than 12.8 — and the microinch equivalents printed on US drawings were rounded again to the familiar 16, 32, 63, 125. The chart above shows both the printed figure and the exact conversion in adjacent columns so it is obvious that the difference is convention rather than error, and that a part measured at 31.8 µin meets a 32 µin callout.

ISO 1302 deprecated the N-grade callout in favor of writing the Ra value on the drawing directly, so a recent print showing N7 has either not been revised or was copied from one that was not. The change is an improvement: a grade number is shorthand for a single Ra maximum and nothing else — N7 means Ra 1.6 µm, not a band around it — and writing 1.6 removes the extra lookup while making room for the parameter and the sampling length beside it. The grades remain useful for reading legacy drawings, which is why the table is here.

Not reliably. Milling holds Ra 0.8 to 6.3 µm as ordinary work, and the low end of that already assumes a sharp finishing pass with light radial engagement, a rigid setup and no chatter. Below Ra 0.8 the answer is a different machine rather than a better pass: grinding covers 0.1 to 1.6 µm, and it is exactly at this threshold that a finish callout stops being free. Chasing a ground finish on a mill usually produces a part that measures acceptably in one place and not in another, because what defeats it is chatter and deflection rather than feed rate.

Because Ra averages the profile and throws away everything about its shape. Lay is not captured at all — a turned face carries a continuous helical groove that leads fluid straight across a seal, while a ground or lapped face of identical Ra does not. Waviness, the longer-wavelength error the roughness filter removes before Ra is computed, is not captured either, and a face that is wavy but locally smooth will not conform to a gasket. Nor are isolated deep scratches, which barely move an arithmetic average and will happily leak on their own. A sealing face, a bearing surface and a fatigue-critical fillet each care about a different feature of the profile, and only one of those features is Ra.

A setup, which is the step that matters rather than the machining time. Ra 1.6 µm on a turned face is a finishing pass on the machine the part is already on. Ra 0.4 µm sits at the extreme floor of what turning reaches and is realistically a grinding operation: a second machine, a second fixture, a second operator and a second inspection. Even staying on the lathe the feed has to fall from about 0.20 to 0.10 mm per revolution at a 0.8 mm nose radius, doubling the cycle time before any of the rest is counted. The cost of a finish specification is a step function, and this is where the step is.

About 0.20 mm per revolution at a 0.8 mm nose radius, from Ra ≈ f² ÷ (32 r). Doubling the nose radius to 1.6 mm lets the feed rise to roughly 0.29 mm/rev for the same finish, which is why a larger radius is the cheapest way to buy surface quality — until the extra radial force starts pushing a slender part away from the tool. Treat the result as a floor rather than a prediction: it is the geometric cusp the insert leaves between passes, and every real effect makes the measured surface rougher than it, never smoother.