CNC Mirror Milled Parts: How a Mirror Finish Is Actually Cut

• 1
  What CNC mirror milled surfaces remove and what they leave
• 2
  Tool geometry and cutting data that make reflection possible
• 3
  Why 5-axis machining holds a mirror better than 3-axis
• 4
  Which metals and plastics take a mirror finish well
• 5
  How a mirror surface is measured and reported
• 6
  When a mirror pass is worth it and when it is not

Milling always leaves a pattern of feed marks. A 12 mm end mill at 0.15 mm per tooth and a 60 percent stepover leaves visible scallops you can feel with a fingernail. CNC mirror milled parts sit at the opposite end of that scale, where scallop height drops below the point your eye can resolve under normal light.

The number that matters is Ra, the arithmetic mean roughness. As-machined milling typically lands around Ra 1.6–3.2 μm. A good high-finish pass reaches Ra 0.8–1.6 μm. The reflective range starts near Ra 0.2–0.8 μm, and that is where the term mirror is used honestly rather than as marketing.

Two things create the reflection. First, the tool nose radius has to be large relative to the stepover, so the overlapping arcs blend into a flat envelope. Second, the cutting edge must shear material instead of rubbing it, because rubbed material tears and smears instead of forming a clean chip.

Mirror milling is a cutting process. It is not polishing, and it is not a coating. If a drawing calls for a mirror surface on a 300 mm stainless housing, that call has to be met by a controlled finishing pass, not by hand work after the fact.

• 1
  Ra 1.6–3.2 μmStandard as-machined finish
• 2
  Ra 0.8–1.6 μmControlled finishing pass
• 3
  Ra 0.2–0.8 μmReflective range, part geometry permitting

The single biggest lever is stepover. Scallop height scales with the square of stepover and inversely with tool radius. Halving stepover cuts scallop height by roughly four times. That is why a mirror pass on aluminum often runs at 0.02–0.05 mm stepover instead of 2 mm.

Tool radius then does the rest. A Ø16 mm bull nose with a 4 mm corner radius tolerates wider stepover than a sharp-cornered cutter before the surface goes hazy. Ball nose tools give the most uniform envelope on curved geometry, but they need a tighter stepover to reach the same Ra.

Speeds and feeds have to stay in the shearing window. In 6061 aluminum that usually means 800–1,500 m/min surface speed and 0.05–0.15 mm per tooth, with the finishing pass taking 0.2–0.5 mm radial and 0.1–0.3 mm axial depth. Push the feed too high and chatter appears; drop the speed too low and built-up edge smears the surface.

Rigidity is not optional. A long tool holder or a thin floor will vibrate, and vibration prints itself into the surface as a periodic pattern. On deep pockets we shorten the gauge length, reduce radial engagement, and sometimes accept a two-pass strategy with a semi-finish at 0.3 mm stock before the mirror pass.

Coolant choice matters more than most shops admit. Aluminum with a high-silicon content and copper alloys respond well to a heavy mist or flood of clean, filtered coolant. Dirty coolant recirculates chips and those chips drag across the finished face.

• 1
  Stepover0.02–0.05 mm on aluminum mirror passes
• 2
  Tool typeBull nose for flats, ball nose for 3D contours
• 3
  Radial depth0.2–0.5 mm, axial 0.1–0.3 mm
• 4
  CoolantFiltered flood or mist, never dirty

A 3-axis machine can only reach a curved surface by interpolating many small line segments. Each segment change is a tiny direction reversal, and the servo has to accelerate and decelerate. The result is a faint faceted pattern that gets worse as the surface tilts away from vertical.

A simultaneous 5-axis machine carries the tool normal to the surface. The lead angle stays constant, so the cutting edge sees the same effective radius across the whole pass. The scallop envelope stays even, and the surface reads as continuous reflection instead of a grid.

The second gain is reach. Undercuts, deep pockets and curved walls often cannot be finished in one setup on a 3-axis machine. Repositioning creates a witness line that no amount of polishing hides cleanly. Five-axis work holds the whole contour in one continuous pass.

The third gain is thermal. Shorter cutting paths and constant engagement mean less heat dumped into the part. On thin-walled aluminum parts, that difference shows up as less distortion, and distortion is what breaks a mirror finish after it has already been cut.

GreatLight runs 16 simultaneous 5-axis machining centers with a Ø400 mm rotary table, alongside 12 four-axis mills and 27 three-axis machines. For a mirror pass on complex geometry, the five-axis route is usually the only one that reaches the required Ra without hand finishing.

Aluminum is the easy case. 6061, 6061-T6, 7075 and 6082 all cut cleanly and hold Ra 0.2–0.8 μm with a sharp carbide cutter. 7075 is slightly more prone to chip welding, so it wants more coolant and a sharper edge. Cast ADC12 has porosity, and pores open up as shiny pits under a mirror pass, so it is a poor choice when appearance matters.

Stainless behaves differently. 303 and 304 work-harden ahead of the cut, so a dull tool rubs instead of shearing and the surface turns dull. 316L is gummier still. 17-4PH in the H1150 condition machines cleanly and accepts a good finish. The rule is simple: fresh edge, positive rake, no dwelling in the cut.

Steel grades like 1018 and 1045 finish reasonably well but tend to show faint feed lines unless stepover is kept tight. 4140 and 4340 in a pre-hardened state above 30 HRC hold a better finish than the same alloy in the annealed state, because hard material shears instead of smearing.

Titanium and Inconel are the hard cases. TC4 (Ti-6Al-4V) and Inconel resist a true mirror because heat concentrates at the edge and the material springs back. They can reach Ra 0.8–1.6 μm with low cutting speed and generous coolant, but pushing toward Ra 0.2 μm is slow and tool-hungry.

Plastics are more forgiving than most people expect. POM and acrylic (PMMA) polish to a very clear surface with a sharp single-flute cutter and air blast. PEEK needs slower speeds to avoid melting. ABS tends to fuzz at the edges, so it rarely reads as a true mirror.

• 1
  Aluminum 6061 / 7075Reaches Ra 0.2–0.8 μm reliably
• 2
  Stainless 303 / 304 / 316LNeeds fresh edge, no dwelling
• 3
  Titanium TC4, InconelRealistic floor around Ra 0.8 μm
• 4
  POM, PMMA, PEEKGood clarity with air blast and light feed

A visual check under a desk lamp is not a measurement. Ra comes off a profilometer trace, and the trace direction matters. A surface can read Ra 0.3 μm across the feed direction and Ra 1.2 μm along it, because the feed marks run one way.

We report Ra with the cutoff length and evaluation length stated, since a 0.8 mm cutoff on a 4 mm evaluation gives a different number than a longer trace. For appearance-critical parts, we also check with a gloss meter or a defined light box, because two surfaces with the same Ra can look different if the waviness differs.

Form tolerance is separate from roughness. A mirror finish on a warped face still reflects a warped image. Where flatness matters, we inspect it on a granite surface plate or with a coordinate measuring machine, not with a roughness gauge.

Every part ships after 100 percent inspection. Raw material is verified on receipt, in-process checks catch drift before the finish pass, and final inspection confirms both dimensions and surface. Reports are available on request, and the shop holds ISO 9001:2015, IATF 16949:2016, ISO 13485:2016 and ISO 27001:2022 certifications.