A buyer once sent back two housings from the same order. Same drawing, same 6061-T6, same "black anodize" line in the notes. One came out charcoal. The other had a brown cast under office lighting. Nothing had been machined wrong, and nothing had been anodized wrong either — the two parts came from different extrusion lots, ran in different anodizing loads eight days apart.

That is the most common surprise in anodized aluminum work, and it is almost always a documentation problem rather than a process failure. Anodizing sits at the end of the chain, after every dimension is already fixed, and it quietly changes three things: the size of the part, the fit of anything threaded or bored, and the color you thought you specified.

The coating grows in both directions

Sulfuric anodize (Type II) builds an oxide layer that does not simply sit on the surface. Roughly half the coating thickness grows outward, half penetrates into the base metal. A 20 µm coating therefore adds about 10 µm to any given surface, and pushes 10 µm into the aluminum.

For a shaft, that means the diameter grows by approximately the full coating thickness, because there are two sides: 20 µm coating, 20 µm larger diameter.

Put that next to a ±0.005 mm tolerance and the arithmetic gets uncomfortable. 20 µm is 0.020 mm. The coating alone is four times the tolerance band, and it moves the part in one direction only — always outward, never back.

So a Ø10.000 ±0.005 mm journal that measures perfectly on the CMM can come back from finishing at Ø10.018 and be scrap. The part was right. The process was right. The sequence was wrong: the tolerance was held before coating instead of after.

Three ways out, in order of how often they are actually used:

Machine undersize on purpose. Leave 15–20 µm of radial stock, let the coater bring the part into the band. This works only if the coater measures and the two of you agree on the target, because coating thickness varies by a few µm across a rack.

Machine after coating. For bearing fits and seal journals this is the normal answer. Anodize oversize, then turn or grind the critical diameter. The coating is hard, roughly 300–500 HV for Type II, so use a sharp carbide insert and light depth of cut or you will chip the edge of the coating instead of cutting it.

Mask the surface. Tape or a lacquered plug keeps the oxide off entirely. Electrical grounding pads, press-fit bores and dowel holes belong here. The catch comes later in this article.

Black is the hardest color to hold

Clear anodize hides a lot. Black does not.

The dye goes into the pores of the coating, so color depth tracks coating thickness and pore structure. A 12 µm coating takes less dye than an 18 µm coating and reads lighter, sometimes with a purple or brown edge under 4000 K light. Two racks that came out at different thicknesses will not match even if they were dyed in the same tank on the same day.

Alloy matters just as much. 6061 with a higher silicon content goes gray rather than deep black after clear anodize. 6063 extrusion has a different grain structure than 6061 bar, and the two dye to visibly different tones. 7075 tends toward dark bronze. A shop that buys bar stock from three suppliers will hand you three shades of black over a year, and every one of them will pass an inspection that only checks "is it coated."

The practical fix is boring but effective: keep one lot of material for one color, run the whole order in one load where the rack size allows, and approve a physical sample rather than a color name. "Black" describes a family, not a value.

If you need to measure it, ΔE is the number. Below 1.0 the difference is not visible to the human eye. Between 1 and 2 a trained eye can catch it side by side. Above 3 it is obvious to anyone, which is where customer complaints start. A hand-held spectrophotometer on a flat, ungrained surface gets you there in a minute per part.

Threads and bores close up

Any hole loses twice the coating thickness in diameter. A Ø3.00 H7 bore with 10 µm per side becomes Ø2.98, and a go/no-go pin that used to enter by hand now needs a press. An M3 internal thread loses roughly 20 µm on the pitch diameter, which is usually enough to make the go gauge bind on the first turn.

The standard answers:

  • Mask internal threads and re-tap after coating if the thread carries load.
  • Specify thread allowance in the drawing, for example 6G instead of 6H, and tell the coater the coating thickness so they can plan for it.
  • For external threads, cut to the low side of the tolerance band.

Blind holes are worse than through holes, because the coating builds slightly thicker near the bottom where the electrolyte circulates less, and masking a deep small bore is unreliable. If a Ø2 mm hole 8 mm deep has to stay Ø2 mm, conversion coating is a better choice than anodize.

Masking leaves a step you can feel

Masked areas do not get coated, so the boundary between coated and uncoated becomes a physical step of about half the coating thickness — 10 µm at a 20 µm coating. On a cosmetic face that step shows up as a faint line under raking light, and on a sealing surface it is a leak path.

Designers who specify "mask the threads and the grounding pad" sometimes forget that the mask line has to land somewhere. Put it on a non-cosmetic face, in a groove, or under a washer, and say so on the drawing.

Type II versus Type III

Type II is the standard sulfuric anodize: 5–25 µm, clear or dyed, takes a seal, keeps dimensional change modest. Most enclosure, heatsink and consumer-facing parts are Type II.

Type III, hard anodize, runs 25–50 µm and reaches 400–600 HV. It wears far better, and it is noticeably more brittle. Sharp corners crack because current density concentrates there and the coating cannot flex. Anything going to Type III should carry a 0.5 mm minimum edge radius, and any thread in a Type III part should be masked without exception.

Hard anodize also darkens the part on its own — the natural color runs from gray to bronze-black depending on alloy, which is why "hard anodize black" is a request worth questioning. You may not need the dye at all, and the dye may not hold.

What to write in the callout

A finish note that produces a repeatable part looks like this:

Anodize per MIL-A-8625 Type II, Class 2, black, 10–15 µm, sealed.
Mask: M3 threads (4x), Ø3 H7 bore, grounding pad marked A.
Color to match approved sample, ref. AN-BLK-02, ΔE ≤ 1.5.

Four things that note does that "black anodize" does not: it names a standard, it fixes a thickness range, it says exactly what stays bare, and it gives the coater something to compare against.

What to check when the box arrives

Eddy-current thickness gauge on five flat points per part, away from edges and corners where the reading always runs high or low. Pin gauges in every masked bore and thread gauge on every masked thread. A visual pass under the same lighting the end customer will use, not under the yellow shop lights. And on dyed parts, one part from each rack against the approved sample.

None of this is exotic. It is the difference between a finish note that describes an intention and one that describes a part.

Aluminum, titanium and PEEK machining, 220 machines, ±0.005 mm, prototypes in 3–7 days. Surface finish options and the tolerances that go with them: https://chxmetal.com/en/surface/