Getting Better Surface Finishes on a CNC Mill

If your parts come off the mill looking like they were chewed instead of machined, the first instinct is usually "this machine is garbage" or "cheap bits." Sometimes true. More often it's stepover, feed per tooth, and the order you try them in — and those are free to fix. This is the list to run through before you blame the hardware.

Most finish problems trace to one of four things: too much stepover on the finish pass, a chip load that's dropped below the tool's minimum, a mismatch between climb direction and your machine's backlash, or runout that's beating up the edge faster than you notice. In that order, because that's the order of impact.

What actually determines surface finish

For a flat end mill — assuming DOC and deflection are in check — the dominant variables are stepover and tram. Spindle speed matters less than most people expect, but it's not irrelevant. Within the correct SFM range for the material, changing speed has modest effect on the ridge pattern. Outside that range, too low invites built-up edge (especially in aluminum), which smears the finish regardless of stepover. Too high generates heat that can affect surface condition in heat-sensitive materials. Get SFM in the ballpark first, then focus on stepover.

Tram is worth understanding before touching the stepover dial. If the spindle isn't perpendicular to the table, every pass leaves a step at the boundary regardless of how fine your stepover is. A tram error of 0.008" (0.2 mm) over 4" (100 mm) — not unusual from the factory on many hobby machines — creates a predictable ridge at each pass edge:

ridge height ≈ stepover × (tram error ÷ measurement span)
             = stepover × (0.008 ÷ 4) = stepover × 0.002

At 0.5" (12.7 mm) stepover, that's a 0.001" (0.025 mm) ridge — detectable with a fingernail, visible under raking light. At 0.25" (6.4 mm) stepover, 0.0005" (0.013 mm) — borderline. At 0.1" (2.5 mm) stepover, 0.0002" (0.005 mm) — about a tenth of a hair's width, below the threshold of touch. Narrowing stepover masks the symptom; tramming eliminates it. If facing passes leave consistent parallel ridges that don't improve with better parameters, check tram before changing anything else.

For ball nose, scallop height between parallel passes depends on effective cutting diameter and tilt. Chip Chad surfaces effective diameter in the recommendation row so you don't have to redo the trig in your head.

The highest-impact change: reduce stepover

If there's one lever that matters most, it's stepover. Halving stepover improves finish more than doubling RPM because you're physically reducing the scallop height, not just polishing it harder. For dedicated finishing passes, machinists commonly target 5–10% of tool diameter on flat end mills and 5–8% on ball nose for 3D surfaces. Chip Chad's auto-tuning targets a wider range — roughly 15% and up — but it won't flag a manually set 5% as wrong, it just won't suggest it on its own. Set WOC manually for finishing passes and Chip Chad compensates feed automatically.

The cycle time tradeoff is real, but finishing passes are a small fraction of total time on most jobs. Paying 30 extra seconds to skip a manual deburr afterwards is almost always worth it.

Climb vs. conventional in finishing

Climb milling leaves a better surface on rigid setups because the chip-thickness profile starts thick and ends thin — the tool exits the cut through air, not through a burr. Conventional milling does the opposite and tends to smear the exit edge.

Exception: if your machine has measurable backlash or flexes audibly, conventional is safer on finishing passes. The climb force direction amplifies slop — the cutter tries to pull the work into itself, and any looseness in the drive system lets it. The common guideline is that climb works well at light depths; deeper cuts on machines with backlash are more reliably run conventional. There's no hard threshold — it's a feel-it-out based on your specific machine.

The chip-load trap at low stepover

As you drop stepover for a better finish, the actual chip thickness drops faster than expected — this is radial chip thinning. If you don't increase feed rate to compensate, the tool ends up rubbing instead of cutting. The result is chatter marks, inconsistent finish, and a surface that somehow got worse as you dialed in finer passes.

Chip Chad compensates automatically when your WOC calls for it — finishing passes and adaptive toolpaths both trigger the adjustment. For the full explanation of why this happens, see Understanding Chip Load.

The finishing-pass recipe

1. Rough to leave 0.005–0.010" (0.13–0.25 mm) of stock on the walls and floor.
2. Finish at the target stepover, climb direction, with chip-thinning-compensated feed.
3. Run one final spring pass at the same numbers — it'll clean up any deflection witness marks.
4. Check with both a fingernail and raking light — they catch different things. Your finger detects ridges your eyes miss under flat lighting. Raking light reveals optical witness marks finer than touch can feel. If it looks bad under raking light but feels smooth, it's usually an optical effect from the tool path rather than a dimensional problem.

If finish is still poor after all that, the problem isn't the parameters. It's tram, rigidity, runout, or a dull tool — check in that order.