NewsNOX METALS Raises an $11.5M Seed Round to Reindustrialize America

Surface Finish Predictor

Enter a feed and a tool nose radius and this returns theoretical Ra and Rz, a realistic achievable band rather than the geometric floor, the maximum feed that still hits the Ra on your print, and scallop height for ball and bull nose work. It also tells you when the callout is not reachable by cutting at all, which is the answer nobody wants and the one worth knowing before you quote.

Last updated July 2026

Operation
Lubrication
Read This Before The Numbers

Theoretical roughness is a geometric floor set by feed marks alone. It ignores built-up edge, tool wear, runout, vibration, chatter and smearing. Real parts come out worse than theoretical, never better. Use the theoretical figure to rule a target out, and the realistic band to plan for it.

Theoretical Roughness, Geometric Floor
Ra
63.9 uin
1.623 um
Rz
255.6 uin
6.492 um
Rz Exact Sagitta
256.6 uin
against the f^2 / 8r form
Limiting Mark
Feed cusp
Rz = f^2 / (8 x r) = 0.0080^2 / (8 x 0.0313) = 2.5559e-4 in = 255.6 uinRa = f^2 / (32 x r) = 0.0080^2 / (32 x 0.0313) = 6.3898e-5 in = 63.9 uin = 1.623 um
Realistic Achievable Ra Band
96 to 147 uin2.426 to 3.733 um
Base Band
1.30x to 2.00x
Material Factor
1.15x (grade C)
letter is a relative chip form and finish ranking, not a spec value, and it does not enter the math
Lubrication Factor
1.00x
Process Floor
8 to 16 uin
multiplier band = 1.30 to 2.00 x 1.15 x 1.00 = 1.49x to 2.30xRa realistic = 63.9 x [1.49, 2.30] = 95.5 to 147.0 uin, no floor applied
The band is a planning range, not a published constant. Its basis: the geometric result ignores every mechanism that adds roughness and none that removes it, so measured finish on a rigid setup with a sharp tool and flood coolant commonly lands 1.3 to 2.0 times theoretical and degrades from there. The process floor is itself a band, because no feed change gets carbide much below its bottom, and once the feed stops being the limit the spread does not collapse either: built-up edge, runout and spindle vibration set the surface at that point instead of the feed, and they scatter wider than the feed math ever does.
Verdict Against The Print
UNLIKELYtarget 63 uin (1.600 um) against a realistic 96 to 147 uin
Max Feed, Geometric
0.0079 in/rev
hits the target on paper only
Max Feed, Realistic
0.0052 in/rev
lands the top of the band on the target
Radius Needed At This Feed
0.0317" geo
0.0730" real
Process For This Target
Standard finish pass. No special tooling needed.
f_max = sqrt(32 x r x Ra_target) = sqrt(32 x 0.0313 x 6.300e-5) = 0.00794 inf_max realistic = sqrt(32 x r x Ra_target / 2.30) = 0.00524 in
What To Change

The geometry alone already exceeds the print. At 0.0080" feed on a 0.0313" radius the feed marks are 63.9 uin before anything goes wrong. Drop the feed to 0.0052" or fit a 0.0730" radius. Nothing about tooling or coolant recovers a geometric shortfall.

Gummy in T6 and T651 and the most built-up-edge prone of the group. BUE welds aluminum to the cutting edge, then it tears off and takes finish with it, which happens whatever the feed is. On the Aluminum Association A to E scale, which rates chip form and achievable finish, the commonly reproduced tables put 6061-T6 and T651 at C against B for 7075, though sources differ by a letter. That is not a speed ranking: 6061 runs faster and pulls less power than 7075, it just does not finish as cleanly.

What Governs This Operation

The feed and nose radius model fits turning and boring best, because one edge makes every mark and the geometry repeats exactly once per revolution.

Best case for aluminum. Flushes chips out of the cut so they are not recut, and keeps the edge cool enough to slow built-up edge.

Gummy in T6 and T651 and the most built-up-edge prone of the group. BUE welds aluminum to the cutting edge, then it tears off and takes finish with it, which happens whatever the feed is. On the Aluminum Association A to E scale, which rates chip form and achievable finish, the commonly reproduced tables put 6061-T6 and T651 at C against B for 7075, though sources differ by a letter. That is not a speed ranking: 6061 runs faster and pulls less power than 7075, it just does not finish as cleanly.

Feed Marks And Where Ra Is Measured
r = 0.0313"tool nose radius, not to scalefeedpeak linemeanRz0.0080"vertical scale exaggerated
Ra is the average deviation from the mean line, Rz is the peak to valley height of the same profile. Both figures above describe the feed marks and nothing else, which is why a part can measure two or three times worse than the geometry says while looking correct under a loupe. Ra and Rz are not interchangeable and no single conversion is exact, the Rz to Ra ratio runs about 4 on a clean turned profile and higher on a torn or smeared one.

How to use this calculator

  1. Pick the operation. Turning and boring fit the feed and nose radius model best because one edge makes every mark. Face milling uses feed per tooth and the insert nose radius. Ball and bull nose milling adds a stepover.
  2. Enter the feed. Per revolution for turning and boring, per tooth for either milling mode. This is the feed at the finish pass, not the roughing feed.
  3. Enter the tool nose radius. A 1/32 inch insert nose is 0.0313 inch, a 1/64 is 0.0156, a 0.8 mm nose is 0.0315. For ball and bull nose work enter the cutter or corner radius and the stepover instead.
  4. Enter the target Ra from the print in microinches. If the print is in micrometres, convert first with the surface finish converter, which handles Ra, Rz, RMS and CLA.
  5. Set the material and the lubrication. These do not change the theoretical number, only the realistic band, because they change how much built-up edge and smearing gets added on top of the feed marks.
  6. Read the verdict. If the theoretical figure alone exceeds the target, the geometry is the problem and no tooling change fixes it, so cut the feed or fit a bigger radius. If only the realistic band exceeds the target, condition is the problem, so look at stickout, runout, coolant and edge sharpness.
  7. Use the max feed at target figures to set the finish pass. The geometric one is the paper answer, the realistic one is the feed to actually program.

Typical achievable surface finish by process

Ranges are typical process capability on aluminum, not guarantees, and they assume a sharp tool and a rigid setup. The Rz column uses the geometric convention Rz is about four times Ra, which holds on a clean turned or milled profile. On a torn or smeared surface the ratio climbs, which is why Ra and Rz are not interchangeable and no single conversion is exact.

ProcessRa (microinches)Ra (micrometres)Rz approx (microinches)Relative cost
Bandsawn cut, as-cut500 to 2,00012.5 to 502,000 to 8,000Included in the cut, no extra operation
Rough milled250 to 5006.3 to 12.51,000 to 2,000Baseline, material removal rate is the goal
Rough turned125 to 2503.2 to 6.3500 to 1,000Baseline
As-milled, standard finish pass63 to 1251.6 to 3.2250 to 500Baseline, what you get when the print has no callout
Finish turned or bored32 to 630.8 to 1.6125 to 250Slight, one extra light pass
Finish milled, sharp tool, light feed32 to 630.8 to 1.6125 to 250Slight, one extra light pass
Reamed32 to 630.8 to 1.6125 to 250Extra tool and an extra operation per hole
Cast tool and jig plate, as-supplied machined face32 to 630.8 to 1.6125 to 250Included in the plate, but porosity pits read worse than the profile
Fine turned, wiper insert, low feed16 to 320.4 to 0.863 to 125Moderate, wiper insert plus a slower finish feed
Fine finish milled, polished flute, short stickout16 to 320.4 to 0.863 to 125Moderate, finishing cycle time roughly doubles
Surface or cylindrical ground8 to 320.2 to 0.832 to 125High, separate machine and setup
Fine ground4 to 160.1 to 0.416 to 63High, frequent dressing and slow infeed
Honed4 to 160.1 to 0.416 to 63High, bores only, dedicated tooling
Single crystal diamond turned1 to 80.025 to 0.24 to 32Very high, dedicated machine and a fragile tool
Lapped1 to 80.025 to 0.24 to 32Very high, per part labor
Polished or superfinished0.5 to 40.012 to 0.12 to 16Highest, per part labor with no dimensional control
Bead blasted, non directional cosmetic60 to 1501.5 to 3.8250 to 600Low, but it hides marks rather than reducing Ra
Type II anodize over a 63 uin base63 to 1251.6 to 3.2250 to 500Anodize does not smooth, it follows and slightly roughens the base

Material and lubrication factors used in the realistic band

The grade column is the Aluminum Association A to E machinability rating, which scores chip form and achievable finish only. It is not a speed or power ranking. 6061 rating a letter below 7075 surprises people, and the reason is exactly the subject of this tool: 6061 in T6 and T651 is gummy and built-up-edge prone, and built-up edge ruins finish no matter what the feed is. 6061 still runs at higher cutting speeds and pulls less power than 7075.

Basis for the grade column: published A to E tables differ by a letter between sources and between tempers of the same alloy, so treat the letter as a relative ranking of chip form and finish, not as a spec value, and do not quote it to a customer as one. The finish factor column is what this calculator actually uses. It is a calibrated planning figure derived from how much built-up edge and smearing each alloy adds on top of the feed marks, and it is set independently of the letter.

Material or conditionMachinability gradeFinish factorNote
6061-T651C1.15xGummy, the most built-up-edge prone of the group, but runs faster and pulls less power
7075-T651B1.00xBest chip form and finish of the group, shorter chips and less smearing
7075-T7351B1.02xOveraged and slightly softer, marginally more smearing than T651
7050-T7451B1.02xThick-section standard, finishes the same at the bottom of a deep pocket as at the top
5000 series cast tool and jig plateC to D1.25xPorosity and hard particles show as pits no feed change removes, and it anodizes blotchy
Flood coolantn/a1.00xBest case, clears chips and slows built-up edge
MQLn/a1.12xChips linger and recut chips leave witness marks
Dryn/a1.30xInvites built-up edge and smearing, worst case on 6xxx

Formulas

All lengths are in inches unless noted. Multiply an inch value by 1,000,000 to read it in microinches, and multiply microinches by 0.0254 to read micrometres.

Rz exact = r - sqrt(r^2 - f^2 / 4)
Rz = f^2 / (8 x r)
Ra = Rz / 4 = f^2 / (32 x r)
f_max at a target = sqrt(32 x r x Ra_target)
r needed at a given feed = f^2 / (32 x Ra_target)
scallop h exact = R - sqrt(R^2 - (ae / 2)^2)
scallop h approx = ae^2 / (8 x R)
feed cusp on a ball nose = fz^2 / (8 x R)
Rz on a ball nose = max(scallop h, feed cusp)
ae_max at a target = 2 x sqrt(2 x R x h - h^2), with h = 4 x Ra_target
multiplier band = 1.3 to 2.0 x material factor x lubrication factor
Ra realistic = Ra theoretical x multiplier band
Ra realistic band floor = process floor to 2 x process floor
Ra [microinches] = Ra [inches] x 1,000,000
Ra [micrometres] = Ra [microinches] x 0.0254

f is feed per revolution for turning and boring, and feed per tooth for milling, both in inches.

fz is feed per tooth, used where the distinction matters.

r is the tool nose or insert corner radius in inches.

R is the ball or bull nose cutter radius in inches, which is half the diameter on a full ball.

ae is the stepover between adjacent passes in inches.

h is scallop or cusp height, the ridge of uncut material left between passes.

Rz is the peak to valley height of the profile, Ra is the average absolute deviation from the mean line of the same profile.

Where the constants come from

Rz is pure geometry. A round tool nose of radius r advancing f per revolution leaves a cusp whose height is the sagitta of a chord of length f on a circle of radius r, which is r minus the square root of r squared minus f squared over four. Expanding the root for f much smaller than r gives the shop form f squared over 8r. On the default case, a 0.008 inch feed on a 0.0313 inch nose, the exact form gives 256.6 microinches and the shop form gives 255.6, a 0.4 percent difference. Truncating the series always drops a positive term, so the shop form is always slightly below the exact height, never above it. The calculator reports both.

Ra is not an independent constant, it follows from Rz. Treat one cusp as a parabola of depth Rz over a pitch f. The mean line then sits two thirds of Rz below the peak line, and integrating the absolute deviation about that mean line gives Ra equal to 0.2566 times Rz, which is f squared over 31.18 times r. The rounded engineering convention Ra equals Rz over 4, that is f squared over 32r, sits 2.6 percent below that derivation, so it is very slightly optimistic rather than conservative. It is still the form used here, because it makes f equals the square root of 32 r times Ra_target an exact inverse that round trips, and 2.6 percent is far inside the 1.3 to 2.0 times real-world band that sits on top of it. Published imperial versions bracket the same ground: one widely used form is Ra in microinches equals 31,675 times f squared over r, against 31,250 for the 1/32 convention and 32,075 for the parabolic derivation, so the whole spread is under 3 percent.

The scallop approximation ae squared over 8R is the same expansion applied to stepover instead of feed. On a 0.250 inch cutter radius at a 0.030 inch stepover the exact form gives 450.4 microinches and the approximation gives 450.0, so the approximation is 0.09 percent low, in the same direction and for the same reason as the feed cusp. At practical stepovers either is fine. The exact form is what the calculator reports.

The realistic multiplier band is a planning range, not a published constant, and it is deliberately shown as a band. Its basis is that every mechanism the geometry ignores adds roughness and none removes it: built-up edge, tool wear, runout, spindle and fixture vibration, chatter and smearing. Measured finish on a rigid setup with a sharp tool and flood coolant commonly lands 1.3 to 2.0 times theoretical, and degrades from there as conditions do. The material and lubrication factors scale that band, and a per process absolute floor is then applied because no feed change gets carbide much below roughly 8 microinches on a turned aluminum surface, 12 on a face mill or 16 on a ball nose. That floor is applied as a band from the floor to twice the floor, not as a single number, because once the feed stops being the limit the surface is set by edge condition and machine dynamics, and those scatter wider than the feed math does.

What this tool does not do

It does not convert between roughness units, use the surface finish converter for Ra, Rz, RMS and CLA. It does not calculate RPM, feed rate or chipload, use the speeds and feeds calculator. It does not model deflection, stickout, length to diameter ratio or chatter, which are the usual reason a finish comes out worse than this predicts, so run the tool deflection and chatter calculator alongside it. And it judges one Ra callout in isolation, so for whether a whole set of dimensional callouts is achievable use the tolerance feasibility checker.

Finish starts with the stock. Consistent temper and known residual stress mean the finish pass cuts the same on the last part as it did on the first, so get an instant quote on 6061, 7075 and 7050 plate cut to size with a mill cert on every order, because a surface finish problem that turns out to be a temper substitution is the most expensive kind.

Frequently Asked Questions

What surface finish can you actually hit when machining aluminum?

A normal finish pass on 6061 or 7075 plate lands between 63 and 125 microinches Ra with no special effort. A deliberate finish pass with a sharp tool and a light feed gets you 32 to 63 microinches. Getting inside 32 microinches takes polished-flute uncoated carbide or a ZrN or TiB2 coating, a high helix, minimum tool stickout and flood coolant, and it roughly doubles cycle time. Below about 16 microinches you are at the practical limit of carbide on aluminum, and below 4 microinches you are in lapping and polishing, not machining.

Is the theoretical surface roughness formula accurate?

It is accurate as a floor and useless as a prediction. Ra = f squared divided by 32 times the nose radius describes the feed marks a round tool nose leaves and nothing else. It ignores built-up edge, tool wear, runout, spindle and fixture vibration, chatter and smearing, every one of which adds roughness and none of which removes it. Real parts measure worse than theoretical, never better. Use the theoretical number to rule a target out, because if the geometry alone exceeds the print no tooling or coolant change will save it, then use a realistic band of roughly 1.3 to 2.0 times theoretical on a rigid setup to plan what you will actually measure.

How do I calculate the feed needed to hit a target Ra?

Rearrange the roughness formula: f equals the square root of 32 times the nose radius times the target Ra, with all three in inches. For a 63 microinch target on a 0.0313 inch nose radius insert that gives 0.0079 inch per revolution. That is the geometric answer. Because real finish runs a multiple of theoretical, divide the target Ra by the top of your realistic multiplier before taking the square root, which on gummy 6061 with flood coolant drops the same case to about 0.0052 inch per revolution.

Does 7075 give a better surface finish than 6061?

On chip form and achievable finish, usually yes. The Aluminum Association machinability scale runs A to E and rates chip form and finish quality only, not speed or power, and in the commonly reproduced tables 7075-T6 and T651 sit at B against C for 6061-T6 and T651. Published tables do vary by a letter between sources and tempers, so treat that as a relative ranking rather than a spec value. The physical reason behind it is not in dispute: 6061 in T6 and T651 is gummy and the most built-up-edge prone of the common plate alloys, and built-up edge wrecks finish independently of feed. That is not the same as saying 7075 machines better overall. 6061 runs at higher cutting speeds and pulls noticeably less power, so 7075 costs you cycle time and spindle load to buy the cleaner surface.

What is scallop height and how do I calculate it for a ball nose cutter?

Scallop or cusp height is the ridge of uncut material left between adjacent passes of a round-bottomed cutter. The exact form is h equals R minus the square root of R squared minus stepover squared over four, where R is the cutter radius. The common approximation h equals stepover squared over 8R agrees within about a tenth of a percent at practical stepovers, so either is fine, and the exact form is what this calculator reports. On a real three dimensional contour the effective radius drops on steep walls, so the same stepover leaves a taller scallop there than it does on a flat floor.

Why is my measured Ra worse than the calculated value?

Because the calculation only covers feed marks. The usual culprits on aluminum, in the order worth checking, are built-up edge on the cutting edge, recut chips because coolant is not clearing the pocket, tool deflection and chatter from too much stickout, spindle or toolholder runout, and insert height scatter on a face mill where one insert cutting proud sets the whole surface. None of these appear in the feed and radius math. If the measured value is two to three times theoretical, that is normal. If it is five times or more, you have a process problem, not a feed problem.

Does anodizing improve surface finish?

No. Anodize follows the surface underneath it and usually reads slightly rougher, because the coating grows into and out of the substrate rather than filling it. A 63 microinch base finish anodizes to roughly 63 to 125 microinches. If a print calls out both a fine Ra and an anodize, the Ra has to be achieved before the tank, not by it. Cast 5000 series tool and jig plate is a special case, because it anodizes blotchy regardless of how well it was machined, so do not plan a cosmetic anodize over it.

Related Tools

Related Reading