First article inspection does not care how fast you cut. It cares whether the surface, the size, and the geometry repeat. Most finishing failures trace back to three parameters set by habit rather than by calculation.
Scallop height is a decision, not a default
Formula
h = R − √(R² − (s/2)²)
Work backwards from the required Ra rather than accepting the CAM default. A 6 mm ball mill at 0.25 mm stepover leaves roughly 2.6 µm of cusp — acceptable for a bracket, unacceptable for a sealing face.
Finishing parameters by material
| Material | Tool | Speed (m/min) | Feed/tooth | Stepover | Typical Ra |
|---|---|---|---|---|---|
| 6061-T6 aluminium | 3F carbide, AlTiN | 300–450 | 0.06 mm | 8% Ø | 0.8 µm |
| 7075-T6 aluminium | 3F carbide, polished | 250–400 | 0.05 mm | 6% Ø | 0.6 µm |
| 304 stainless | 4F carbide, coated | 80–120 | 0.04 mm | 5% Ø | 1.2 µm |
| Ti-6Al-4V | 5F carbide, high helix | 40–70 | 0.03 mm | 4% Ø | 1.6 µm |
| Acetal (POM) | 2F uncoated | 400–600 | 0.10 mm | 10% Ø | 0.9 µm |
Constant engagement beats high feed
- Trochoidal and adaptive paths hold radial engagement steady, so deflection stays constant and the wall stays straight.
- Corners are where engagement spikes — let the CAM slow the feed rather than the operator slowing the whole program.
- Climb mill finishing passes without exception; conventional milling burnishes and work-hardens.
gcode
G0 X-12.500 Y0.000 Z5.000
G1 Z-8.000 F400.
G3 X0.000 Y-12.500 I12.500 J0.000 F1200. ; arc lead-in
G1 X40.000 F1600. ; finishing contour
G3 X52.500 Y0.000 I0.000 J12.500 ; arc lead-out
G0 Z25.000Before you press cycle start
- Tool stickout measured and entered, not estimated
- Semi-finish stock uniform on walls and floors
- Lead-in and lead-out arcs on every inspected surface
- Coolant reaching the cut, not the shroud
- Probe cycle confirming datum before the finishing operation
Frequently asked
Ball mill or bull nose for floors?
Bull nose. It leaves a flatter floor with a larger effective cutting diameter and far less cusp.
How much does stickout really matter?
Deflection scales with the cube of stickout. Reducing it by 20% roughly halves deflection.
