Surface Finish on Aluminum: What Ra You Can Actually Hit, and What the Print Is Really Asking For

Ra Is an Average, and That Is the Whole Problem

Ra is the arithmetic mean deviation of the roughness profile from its mean line, taken over a defined length after form and waviness are filtered out. Because it averages absolute values, it is deliberately blind to individual features. One deep drag mark from a recut chip barely moves it, and a surface that is uniformly hazy from smearing can read the same as one that is optically bright with a gouge in it. That blindness is why Ra became the universal number and why it rarely captures what the designer was actually worried about. To move between Ra, Rz, RMS, and CLA units, use /resources/surface-finish-converter. To ask what a given tool and feed will actually produce, a different question entirely, use /resources/surface-finish-predictor.

The Geometry That Sets the Floor

In turning, the tool nose is an arc of radius r stepped along the workpiece by the feed f once per revolution. What it leaves is a row of scallops whose height is pure geometry: no material properties, no tool condition, no spindle. Just a circle and a step. The exact sagitta of that scallop is r minus the square root of r squared minus f squared over 4, but nobody uses the exact form, because for any sane ratio of feed to nose radius the approximation below is within a fraction of a percent and far easier to invert.

Formula One: Peak to Valley Height

Formula Two: Turning That Into Ra

For a periodic profile of near-triangular scallops the arithmetic average works out to about a quarter of the peak-to-valley height. A caution on the constant, because sources disagree. Integrate the parabolic approximation to the arc properly and you get f squared over 18 root 3 times r, a constant of 31.2. The widely quoted engineering form rounds it to 32 so Ra is exactly Rz over 4. One published imperial version uses Ra in microinches equals 31,675 times f squared over r, a constant of 31.6. All three give 64 to 66 microinches for the same cut, and that 3 percent spread is noise next to the real-world effects below, so use 32.

Inverting It: the Fastest Feed That Could Still Make the Number

The useful direction is backwards: given a target Ra and a nose radius, solve for feed. Two things fall out of the algebra. Feed enters as a square root, so halving the required Ra costs you only 29 percent of your feed rate rather than half of it. And a bigger nose radius buys feed for free, since doubling r allows 1.41 times the feed at the same theoretical Ra. That is why a large-radius or wiper insert is the standard move when cycle time and finish are fighting.

Why the Measured Number Comes In Higher

Measured Ra is essentially always above theoretical, and the gap is not a constant you can multiply by. Published turning studies and our own shop data agree on the shape of the relationship and disagree on the magnitude, which is itself the finding. In the well-behaved region, meaning a sharp insert, a rigid setup, and feed above roughly 0.006 in/rev, measured Ra commonly lands within about 1.0 to 1.5 times theoretical. As feed drops the ratio climbs, and below roughly 0.003 to 0.004 in/rev ratios of 2 to 5 times theoretical are routinely reported, because the geometric term shrinks toward zero while edge radius plowing, built-up edge, and vibration do not. Anyone handing you a single multiplier is hiding the interesting part. The only way to know yours is to cut a test part on your machine with your tooling and measure it.

Achievable Ra by Process

This is the table people actually need. The bands are what each process delivers in aluminum in ordinary production, not the hero number from a trade show demo. Best with real effort is what a shop that cares can hit with dedicated tooling, a light dedicated pass, and inspection to prove it.

The Finish Ladder, and What Each Rung Costs

125 Ra versus 63 Ra is such a common argument because the two sit on opposite sides of a process boundary. 125 uin, or 3.2 um, is what a competent finish pass produces without anyone thinking about it, which is why it appears as a blanket drawing note everywhere and why it should carry no upcharge. 63 uin generally means one more deliberate pass with a tool reserved for finishing. 32 uin means controlling tool condition, runout, and deflection on purpose and verifying with a profilometer, and that is the rung where quoted cost moves noticeably. 16 uin and below usually changes the process outright.

In Milling, the Feed Term Is Almost Irrelevant

Turning at least gives the feed term a fighting chance, because feed per revolution is large relative to the nose radius. Milling does not. The cusp between successive tooth marks is governed by the cutter radius, which is enormous next to an insert nose, so the scallops collapse to almost nothing.

The Aluminum Problem: Built-Up Edge

Aluminum has a low melting point, a high affinity for tool materials, and enough ductility to smear rather than fracture, so the chip pressure-welds itself to the rake face just behind the cutting edge. That deposit, built-up edge, is not a wear mode, it is a geometry change: it blunts the edge, throws the effective rake away from what you selected, and periodically breaks off and drags across the surface you just finished. A part that starts at 40 uin and drifts to 90 uin across a run with no program change is almost always a BUE story. Two conditions drive it, and both run against instinct. BUE is worst at low cutting speed, because the interface never gets hot enough to stop material sticking, which is why aluminum tooling runs 800 to 3000 SFM instead of backing off. And it is worst at very low chip thickness, because an edge with a 0.0002 to 0.0004 in edge radius cannot shear a chip thinner than itself. It plows and smears instead.

Machinability Letters: What B and C Actually Mean

The Aluminum Association grades wrought alloys on a machinability letter scale from A to E, and the criteria are chip form and achievable surface finish. Not cutting speed, not spindle power, not tool life. On that scale 7075-T6 and T651 are rated B and 6061-T6 and T651 are rated C, which is the source of the trade's most misquoted line, that 7075 machines better than 6061. What the letters actually say is narrower and more useful: 7075 breaks its chips more willingly and gives up a better finish for less effort, because at roughly 150 HB it shears cleanly where 6061 at roughly 95 HB deforms and smears. Meanwhile 6061 tolerates higher surface speed and pulls less cutting force and spindle power at the same chip area, precisely because it is softer. So 7075 grades better on chip form and finish, and 6061 runs faster and lighter. Which one machines better depends on which you are short of: if the constraint is a 32 uin callout, the 7xxx plate is your friend, and if it is a small spindle and a cycle time target, 6061 is.

What Actually Fixes Finish on Aluminum, in Order

This is deliberately the reverse of what most people try first. Everything at the top is free or nearly free, and feed math sits at the bottom because by the time it matters, everything else is already right.

Budget the Number, Do Not Hope for It

Plan a tight finish callout the way you plan a tight dimensional stack: give every contributor an allowance and add them up before you cut. A 32 uin target is not 32 microinches of feed marks, it is 32 microinches shared across everything on the trace, with something left for the tool going dull between the first part and the last.

Ra to Rz Is Not a Ratio

The theoretical model makes Ra exactly Rz over 4, and that clean relationship is where the folklore comes from. Real surfaces do not cooperate. Rz is a peak-to-valley measure, so it is dominated by the extremes that Ra was designed to average away. On a clean periodic turned profile the ratio does sit near 4. On a torn, smeared, or abrasively finished surface, 6 to 10 is ordinary, and one deep isolated scratch can push Rz past 15 times Ra while Ra barely moves. That is not measurement error, it is two parameters correctly reporting different things.

What Anodizing Does to the Number

Anodize does not smooth anything. It converts the outer skin of the part to oxide, replicating whatever was underneath and adding its own texture, so Ra measured after anodize is at or above what you measured before. Type II sulfuric coatings typically run 0.0002 to 0.001 in total thickness with roughly half growing outward from the original surface, and Type III hardcoat is thicker and more crystalline, so its effect on roughness is larger. The pre-treatment is often the real story: a caustic etch for a matte appearance attacks preferentially and can move a machined surface up by a large factor on its own. Three consequences follow. Anodize makes tool marks more visible, not less, because it strips the light-scattering smear and lays a uniform tint over the remaining topography. Growth is real and belongs in your pre-anodize target, which is what /resources/anodize-growth-calculator is for. And an Ra callout on an anodized part is ambiguous unless the drawing says whether it applies before or after coating. Most do not say. Ask in writing before you cut, because the two requirements can be a full ladder rung apart.

What the Print Is Really Asking For

Finish callouts are usually a proxy for a functional worry, and the proxy is often the wrong parameter for the worry. Reading through to the intent is what lets you push back intelligently instead of either eating the cost or shipping a part that passes inspection and fails in service.

• A sealing face is worried about leak paths, which are peaks, valleys, and lay, not an average. Rz or Rmax plus a specified lay direction says it properly. A blanket Ra can pass with a radial scratch straight across the seal.
• A fatigue-critical surface is worried about crack initiation sites, meaning deep isolated features. That is an Rz or Rmax concern, and it is why 7xxx parts in sustained tension deserve a peak-to-valley limit and not only an Ra number.
• A cosmetic surface is worried about how uniform it looks, which Ra does not measure at all. Two surfaces at identical Ra look completely different if the lay differs, and anodize amplifies the difference.
• A bonded or adhesive joint usually wants more roughness, not less. A tight Ra maximum on a bond face is often a copied note working against the design.
• A press fit or bearing bore is worried about the fit changing as high spots wear in. That is a bearing-ratio question, handled with a finish spec plus a fit callout, which is where /resources/tolerance-feasibility-checker earns its keep.
• A blanket 125 Ra note applied to every surface, including sawn edges and drilled hole bottoms, is a template rather than a requirement. Say so early, get a deviation on the surfaces that do not matter, and spend the money on the one that does.