Hard Anodizing Dimensional Control: Why Your Bearing Bores Fail After Coating

Here is a failure that happens somewhere in robotics manufacturing every week.

A joint housing is machined. The bearing bore measures 50.008 mm — comfortably inside its H6 band. First article inspection passes. The parts ship to a finisher with a drawing note reading “Type III hard anodize all over.” They come back looking excellent: uniform, dark grey, exactly what the designer imagined.

Then assembly tries to fit the bearing, and it won’t go. The bore now measures around 49.96 mm. It’s 0.05 mm undersize — more than three times the entire H6 tolerance band, on a feature that was in specification when it left the machine.

Nobody did anything wrong. The machine shop machined to print. The finisher coated to specification. The drawing was internally consistent. And the parts are scrap.

This article is about why that happens, exactly how much dimensional change to expect, and the four ways to prevent it — along with a second, less obvious reason to keep hard coating off your precision features that has nothing to do with dimensions at all.

The Failure Mode: What Actually Happens to Your Parts

A Concrete Example: The 50 mm Bearing Bore

Work the numbers on the case above, because the scale of the mismatch is what makes this failure so consistent.

A 50 mm H6 bore has a tolerance band of +0.016 / 0 mm. That’s 16 micrometres of total permissible variation — the machining tolerance you’d hold for a cross-roller bearing seat in a robot joint.

Type III hard anodizing per MIL-A-8625F defaults to a coating thickness of 0.002 inches, or approximately 50 micrometres, unless the drawing specifies otherwise. Because anodizing both builds outward and penetrates inward, the diameter of an internal bore reduces by approximately the full coating thickness.

So a 50 μm coating takes roughly 50 μm out of the bore diameter. Against a 16 μm tolerance band.

The coating doesn’t nudge the part to the edge of tolerance. It moves it three tolerance bands past the lower limit. There is no machining strategy, no careful process control, and no inspection regime that saves this part — the specification itself was impossible.

Why This Gets Discovered Late

The reason this failure is so expensive is timing. It gets caught at assembly, which is the worst possible place.

The machining inspection passes, because the part was correct when measured. The finishing inspection passes too, because finishers verify coating thickness, adhesion, and appearance — not the customer’s assembly fits. Incoming inspection at the customer, if it happens at all, usually samples cosmetic and gross dimensional attributes rather than re-verifying every H6 bore.

So the discovery point is a technician on the assembly bench discovering a bearing won’t press in. By then the entire batch is coated, the lead time is consumed, and the recovery options are all bad: scrap and remake, or attempt to machine the coating out.

The Blame Cycle Between Machine Shop and Finisher

What follows is predictable and unproductive.

The machine shop produces its inspection report showing the bore was in tolerance. The finisher produces its certification showing coating thickness within specification. Both are telling the truth. Neither is responsible for the interaction between them, because nobody was ever assigned that responsibility.

This is the structural problem underneath the technical one. Coating-induced dimensional change is an interface failure, not a process failure. It happens in the gap between two suppliers who each did their job correctly, and it keeps happening because most supply chains have no one who owns that gap.

Everything else in this article is about closing it.

The Physics: Why Anodizing Changes Dimensions

Conversion vs. Deposition — Anodizing Isn’t Plating

The critical distinction, and the one that trips up engineers who learned dimensional compensation on plated parts.

Plating is deposition. Electroless nickel, hard chrome, electroplated zinc — these add material on top of the existing surface. The original surface stays where it was; a layer sits on top of it. A 25 μm nickel plate on a bore reduces the diameter by 50 μm, because 25 μm is added to each side.

Anodizing is conversion. The process electrochemically converts the aluminium substrate itself into aluminium oxide. Material isn’t added — existing material is transformed. Because aluminium oxide occupies more volume than the aluminium it consumed, the converted layer expands, growing both outward from the original surface and inward into the part.

This is why anodizing has a growth ratio rather than a simple additive thickness, and why the compensation arithmetic differs from plating.

The 50/50 Growth Rule and What It Means for Diameters

The working rule for hard anodizing is that the coating grows approximately 50% outward and 50% inward from the original surface plane. A 50 μm coating means the finished surface sits about 25 μm above where the original surface was, and the coating extends about 25 μm below it into what used to be substrate.

This ratio is an approximation. It varies somewhat with alloy, bath chemistry, current density, and coating thickness, and different sources cite ratios from roughly 50/50 to 60/40. For design purposes, 50/50 is the standard assumption and it’s close enough — the variation it introduces is smaller than the coating thickness tolerance you’ll be dealing with anyway.

The consequence for diameters follows directly. Buildup is 50% of coating thickness on each surface, and a diameter has two surfaces:

• External diameters increase by approximately the coating thickness (2 × 0.5T)
• Internal diameters decrease by approximately the coating thickness (2 × 0.5T)
• Flat surfaces move outward by approximately half the coating thickness (0.5T)
• A slot or groove width decreases by approximately the coating thickness

The rule worth memorising: diameter change ≈ coating thickness. Not half of it, not twice it. For a rough field check that’s all you need.

Coating Thickness by Type and Specification

Coating thickness varies by process type, and the defaults matter because they’re what you get when the drawing doesn’t specify.

Process	Typical thickness	Diameter change (bore)	Common default
Type I (chromic acid)	0.5-7 μm	0.5-7 μm	Rarely specified for robotics
Type II (sulphuric, decorative)	5-25 μm	5-25 μm	~10-15 μm typical
Type III (hard coat)	25-75 μm	25-75 μm	~50 μm (0.002 in) per MIL-A-8625F
MAO / PEO	20-80 μm typical	20-80 μm	Varies widely by supplier

Two things to take from this table.

First, even Type II can break a precision fit. A 15 μm Type II coating consumes the entire H6 band on a bore between 18 and 30 mm. Designers who know to worry about hard coat often assume decorative anodising is dimensionally harmless. It isn’t.

Second, the thickness itself carries a tolerance. Finishers typically hold coating thickness to around ±20% of nominal unless a tighter band is specified and paid for. On a 50 μm nominal coating, that’s an actual range of 40 to 60 μm — meaning 20 μm of variation in the finished bore diameter from coating thickness alone, before any machining variation. That 20 μm exceeds the H6 band on most bore sizes.

This is the point that makes pre-compensation harder than it looks, and we’ll come back to it.

Comparison: How Other Coatings Behave Dimensionally

Anodising isn’t the only coating on a robot component, and the dimensional behaviour of the alternatives is worth knowing because it drives process selection.

Coating	Mechanism	Typical thickness	Bore diameter change	Notes
Type II anodize	Conversion	5-25 μm	−5 to −25 μm	Colour-friendly, moderate wear resistance
Type III anodize	Conversion	25-75 μm	−25 to −75 μm	500-700 HV, high wear resistance
MAO / PEO	Conversion	20-80 μm	−20 to −80 μm	Up to ~1500 HV, rougher as-deposited
Electroless nickel	Deposition	5-50 μm	−10 to −100 μm	Uniform in blind features; 2× thickness effect
Hard chrome	Deposition	10-100 μm	−20 to −200 μm	Poor throwing power, uneven in bores
Chromate conversion	Conversion	<1 μm	Negligible	Conductive; the go-to for masked areas
DLC	Deposition	1-5 μm	−2 to −10 μm	Very low friction, minimal dimensional impact
PVD (TiN, CrN)	Deposition	1-5 μm	−2 to −10 μm	Deposition temperature can affect temper
Dry film lubricant	Deposition	5-20 μm	−10 to −40 μm	Often overlooked as a dimensional factor

The pattern to notice: deposition coatings hit diameters twice as hard as conversion coatings of the same thickness, because the full thickness is added to each surface rather than half of it. Engineers who compensate correctly for anodising sometimes under-compensate for electroless nickel by a factor of two.

And note where chromate conversion sits. At under a micrometre, it’s dimensionally invisible. That property — combined with the fact that it stays electrically conductive — is why it’s the standard answer for the areas you mask.

Where Dimensional Change Bites Hardest

Bearing Bores and Precision Fits

The headline case, and the one that generates most of the scrap.

Precision bearing seats in robot joints typically run H6, sometimes H5 on the tightest applications. As covered in our guide to robot joint components, a cross-roller bearing seat commonly requires bore tolerance of ±5 μm with cylindricity within 5 μm.

Set that against coating-induced change:

Bore size	H6 band	H7 band	Type II change (15 μm)	Type III change (50 μm)
18-30 mm	13 μm	21 μm	Exceeds H6	3.8× H6 band
30-50 mm	16 μm	25 μm	Approximately H6 band	3.1× H6 band
50-80 mm	19 μm	30 μm	Within H6 band	2.6× H6 band
80-120 mm	22 μm	35 μm	Within H6 band	2.3× H6 band

There is no bore size at which Type III hard coating is compatible with an H6 fit without compensation or masking. None. If your drawing has both an H6 bore and an unqualified “hard anodize all over” note, the drawing is self-contradictory.

Threaded Features: The 4× Amplification

Threads are worse than plain bores, and this catches even experienced engineers.

Because of the 60° thread flank geometry, coating on the flanks affects the pitch diameter far more than it affects a plain diameter. The standard plating rule applies to anodising as well: pitch diameter changes by approximately four times the radial buildup.

Run the numbers on a 50 μm Type III coating. Radial buildup is about 25 μm. Pitch diameter change is therefore around 100 μm — a full 0.1 mm.

For an M6 × 1.0 internal thread, the 6H pitch diameter tolerance is roughly 140 μm. A 100 μm change consumes about 70% of the entire tolerance band. For the miniature threads common in dexterous hand assemblies — M1.6, M2.0, M2.5 — the tolerance bands are far smaller and the coating simply closes the thread.

Practical guidance: mask threaded features, or tap them after coating. Attempting to pre-compensate a thread for anodising is possible in principle but requires special-order taps and tight coating thickness control, and it rarely justifies the effort compared to masking.

Sealing Grooves and O-Ring Compression

A subtler failure, and one that passes assembly and fails in the field.

An O-ring groove has three dimensions that matter: width, depth, and the diameter it sits on. Anodising changes all three. Groove width narrows by approximately the coating thickness. Groove depth reduces by about half the coating thickness on the groove floor — but the land surfaces on either side also grow outward by half the coating thickness, so effective groove depth relative to the sealing face reduces by roughly the full coating thickness.

The result is increased O-ring compression. Typical O-ring designs target 15-30% compression. On a shallow groove — say 1.5 mm deep for a 2 mm cord — a 50 μm reduction in effective depth adds around 2.5 percentage points of compression. That may be acceptable. Stack it with machining tolerance in the unfavourable direction and coating thickness at the high end of its band, and you can push compression past the material’s limit.

Over-compressed elastomer takes a permanent set, loses its recovery, and the seal degrades — typically after weeks or months of thermal cycling rather than immediately. This is the anodising failure that shows up as warranty returns rather than assembly scrap.

If your enclosure has an IP rating requirement, groove geometry deserves the same masking-or-compensation decision as a bearing bore. We covered the broader sealing-geometry question in the humanoid robot shells guide.

Sliding and Mating Interfaces

Any surface that slides, indexes, or locates against another part is affected, and the effects stack in ways single-feature analysis misses.

A dovetail slide coated on both mating faces loses clearance equal to roughly twice the coating thickness. A pair of dowel-pin holes coated on their bores lose diameter and gain interference on the pins. A flanged interface where both faces are coated gains roughly one full coating thickness in stack height — which, across a multi-part assembly, moves a downstream feature by a visible amount.

The rule of thumb for assemblies: count the coated interfaces in your stack-up and add half a coating thickness for each coated surface in the path. On a four-interface stack with 50 μm coating, that’s 100 μm of accumulated change that no individual part drawing predicts.

Surface Roughness Changes Independent of Dimension

Even where dimensional change is tolerable, surface texture change may not be.

Anodising generally increases surface roughness, and hard coating increases it more than decorative. A surface machined to Ra 0.4 μm may come back from Type III at Ra 0.8-1.2 μm. The coating grows with a somewhat columnar, porous structure that doesn’t replicate the underlying finish exactly.

This matters independently of size for two features in particular. Sealing lands, where surface texture directly affects sealing performance and where a rougher surface provides leak paths. And low-friction sliding surfaces — the tendon guides and cable channels in dexterous hands, where the specification of Ra 0.2 μm exists precisely because a rougher surface abrades the tendon.

For those features, the question isn’t only “will it still fit” but “will it still perform.” Coating a surface whose function depends on smoothness is often the wrong call even when the dimensional change is within tolerance.

Thickness Isn’t Uniform: The Geometry Effects Nobody Specs

Everything above assumes coating thickness is uniform across the part. It isn’t. The variation is geometry-driven, predictable in direction, and almost never appears on a drawing.

Deep Bores and Blind Holes

Anodising thickness depends on local current density and electrolyte exchange. Both degrade with depth into a bore.

As a working guide, coating thickness starts falling off noticeably at depths beyond roughly one bore diameter, and in blind holes deeper than about two diameters, the bottom region may receive only a fraction of the nominal thickness — sometimes half or less. Blind holes are worse than through holes because electrolyte circulation is poorer.

This cuts both ways. If you’re relying on the coating for wear resistance deep in a bore, you may not be getting it. If you pre-compensated a deep bore for full nominal thickness, the deep end will finish oversize while the entry finishes correctly — producing a tapered bore that measures in tolerance at one end and out at the other.

For any bore deeper than about 1.5× its diameter where coating matters dimensionally or functionally, this needs to be discussed with the finisher before the parts are made rather than discovered afterward.

Internal Corners and Radii

Tight internal corners see reduced current density and therefore reduced coating thickness. A sharp internal corner or a very small fillet radius may receive substantially less coating than adjacent flat surfaces.

Where hard coat is being applied for wear resistance, internal corners are the likely first failure point. Where it’s applied for corrosion protection, they’re the likely first corrosion site. Generous internal radii help — another reason beyond machinability to avoid sharp internal corners in coated parts.

External Edges and Sharp Corners

The opposite problem. External edges and sharp corners see elevated current density and grow thicker coating than adjacent surfaces. Because the coating is brittle, thick coating on a sharp edge is prone to chipping and can create a raised, fragile lip.

MIL-A-8625 addresses this by requiring a minimum edge radius for hard coating — commonly cited as about 0.8 mm (1/32 inch) on edges to be hard coated. Sharper edges than that risk coating breakdown.

Practical rule: break every external edge on a hard-coated part to at least 0.5 mm, ideally 0.8 mm. This isn’t cosmetic deburring; it’s a coating quality requirement.

Racking Points and Contact Marks

Anodising requires electrical contact with the part throughout the process. Wherever the rack touches, no coating forms — leaving small uncoated marks, typically a millimetre or two across.

Those marks have to go somewhere. If you don’t specify where, the finisher will choose based on what’s convenient for racking, which may be your sealing face, your cosmetic surface, or your precision bore.

For any part where the coating matters functionally, specify permissible racking zones on the drawing. It costs nothing at design time and eliminates an entire category of surprise.

Mask Boundary Transition Zones

Masking doesn’t produce a perfectly sharp coating edge. There’s a transition zone at the mask boundary — typically one to three millimetres wide — where coating thickness tapers from nominal to zero. Depending on masking method and how well the mask seals, there may also be minor coating penetration under the mask edge.

The design consequence: position mask boundaries at least 2-3 mm away from any feature that needs to be either fully coated or fully uncoated. A mask that ends exactly at the edge of a bearing bore will leave a partially coated band inside the bore. A mask positioned 3 mm back into a relief groove leaves the bore clean.

This is a feature you can design for — a small relief or undercut at the mask boundary gives the finisher a defined place to end the mask and gives you a predictable result.

The Fatigue Penalty: A Second Reason to Mask

Dimensional change is the failure everybody eventually learns about. This one is less discussed and, on cyclically loaded robot parts, potentially more serious.

Why Hard Coat Reduces Fatigue Strength

Aluminium oxide is hard and brittle. The aluminium underneath is comparatively soft and ductile. Under cyclic loading, the brittle coating cracks long before the substrate would, and each of those cracks becomes a stress concentration and a crack initiation site in the substrate beneath.

The magnitude is significant. Reductions in fatigue endurance limit of roughly 30-60% are commonly reported for hard anodised aluminium relative to bare material, with thicker coatings producing larger penalties. The effect is well enough established that MIL-A-8625 itself carries a caution about fatigue reduction on parts subject to cyclic loading.

Type II decorative coatings carry a smaller penalty, roughly in proportion to their reduced thickness — but the effect is not zero.

What This Means for Cyclically Loaded Robot Parts

Robot components are close to a worst case for this effect. A humanoid joint housing may see millions of load reversals over its service life. A finger phalange in a dexterous hand sees a load cycle on every grasp. Leg and ankle structures see impact loading on every step.

If you’ve sized a structural aluminium part against fatigue life and then specified hard anodising over the load path, your actual fatigue life may be substantially below your calculation. The dimensional problem produces scrap at assembly, which is expensive but visible. The fatigue problem produces field failures at some point in the future, which is worse.

The exposure is highest where hard coat is applied for reasons that have nothing to do with the loaded surface — a housing coated all over for corrosion resistance or appearance, where the coating happens to cover a fillet that carries cyclic bending stress.

When to Accept It and When to Design Around It

Hard coat is not disqualified on structural parts. It’s a trade-off that should be made deliberately.

Reasonable to accept where the part is stiffness-driven rather than fatigue-driven, where the coated surface isn’t in the primary load path, where fatigue margins are large, or where wear resistance is genuinely the governing requirement.

Worth designing around where the part is fatigue-critical, where the coating covers a high-stress fillet or radius, where the part is thin-walled and the coating represents a meaningful fraction of section thickness, or where service life requirements are long and inspection access is poor.

The mitigations available:

• Mask high-stress regions — the same masking that protects precision features protects fatigue-critical ones
• Reduce coating thickness — specify 25 μm rather than the 50 μm default where wear demands allow; the fatigue penalty scales with thickness
• Shot peen before anodising — introduces compressive residual stress in the surface layer that partially offsets the coating’s effect, a well-established aerospace practice
• Use Type II instead of Type III where the requirement is corrosion protection rather than wear
• Increase fillet radii in coated regions to reduce the underlying stress concentration

Four Solutions, and When Each Is Correct

There are exactly four ways to deal with coating-induced dimensional change. Each is right in some circumstances and wrong in others.

Option 1 — Mask the Feature

Physically prevent coating from forming on the critical feature. Methods include silicone or rubber plugs for bores, threaded plugs for tapped holes, masking tape for flat regions, and lacquer or wax for irregular geometry.

Advantages: Complete dimensional certainty — a masked feature doesn’t change at all. Avoids the fatigue penalty locally. Preserves the machined surface finish. Works for threads, where compensation is impractical.

Disadvantages: Adds per-part labour cost, which scales linearly with quantity. Requires a clear drawing specification. The masked surface is left bare aluminium unless separately protected, which may be a corrosion concern. Transition zones require design clearance. Masking reliability varies — plugs can seep, tape can lift.

Best for: Bearing bores, threaded features, sealing lands, optical and sensor interfaces, and fatigue-critical regions. This is the default answer for precision features and should be the starting assumption unless there’s a reason to do something else.

Practical note: If a masked feature will corrode in service, chromate conversion coating can be applied to masked areas afterward. At under a micrometre it’s dimensionally negligible and it maintains electrical conductivity — which, as covered in the shells article, also solves the EMI grounding conflict.

Option 2 — Pre-Compensate the Machined Dimension

Machine the feature deliberately oversize by the expected coating growth, so it lands in tolerance after coating.

Advantages: No masking cost. Uniform coating across the whole part. Simple in concept.

Disadvantages: This is where the coating thickness tolerance becomes the governing problem. At a typical ±20% thickness band on a 50 μm nominal coating, the finished dimension carries ±10 μm of coating-induced variation on top of your machining variation. For an H6 fit with a 16 μm band, that’s unworkable. Even at a tightly-controlled ±10%, you’re consuming a third of the band before machining.

It also requires the finisher to actually measure and hold thickness on your part rather than running to a general shop standard — which means a specified thickness tolerance, an agreed measurement location, and usually a higher price.

Best for: Features with tolerances of H8 or looser, larger-diameter features where the tolerance band is proportionally wider, and production runs large enough to justify establishing tight process control with a specific finisher.

Not suitable for: H7 or tighter fits, threads, or any programme where the finisher may change between batches.

Option 3 — Machine After Coating

Coat the whole part, then machine the critical feature to final size, cutting through the coating.

Advantages: Final dimension is fully under machining control. No masking specification needed. No coating thickness tolerance in the stack.

Disadvantages: The coating is 500-700 HV — harder than most tooling. Cutting it requires diamond or CBN tooling, or honing, and tool life is poor. It adds a second machining setup after finishing, with the fixturing and alignment challenges that implies. Coating dust from machining aluminium oxide is abrasive and needs handling.

And the logical objection: if you’re going to remove the coating from the feature anyway, the feature ends up bare aluminium — the same result as masking, at higher cost.

Best for: Situations where masking is geometrically impractical, where a very tight tolerance must be held and no other route achieves it, or where the part was already coated before the problem was discovered and you’re salvaging a batch.

Realistically: This is usually a recovery strategy rather than a planned process.

Option 4 — Don’t Coat That Feature at All

Reconsider whether the feature needs coating, or whether the part needs hard coating at all.

The questions worth asking:

• Is the bore actually a wear surface? A bearing seat holds a press-fit outer race that doesn’t move relative to it. There’s no sliding wear to protect against.
• Is the requirement wear resistance or corrosion resistance? If corrosion, Type II or chromate conversion may be sufficient at a fraction of the dimensional impact.
• Is the coating specified for a reason, or inherited from a previous drawing?
• Would a thinner coating meet the requirement? 25 μm halves the dimensional change and reduces the fatigue penalty.
• Would a different coating serve better? DLC gives superior friction performance at 1-5 μm — dimensionally almost free.

In our experience reviewing robot component drawings, a meaningful proportion of “hard anodize all over” callouts don’t survive this line of questioning. The note is often inherited from a template or a legacy part rather than derived from a requirement, and the surfaces that actually need wear protection are a small subset of the part.

Decision Matrix

Feature	Tolerance	Recommended approach	Why
Bearing bore	H5-H7	Mask	Compensation can’t hold the band
Bearing bore	H8-H9	Compensate or mask	Compensation viable if thickness controlled
Internal thread	Any	Mask, or tap after coating	4× amplification makes compensation impractical
O-ring groove	Sealing-critical	Mask	Compression stack is unforgiving
Sensor / optical interface	±0.1 mm or tighter	Mask	Position tolerance can’t absorb coating
Sliding wear surface	Functional	Coat — this is what hard coat is for	Wear resistance is the requirement
Cosmetic exterior	±0.2 mm or looser	Coat	Dimensional change is negligible at this tolerance
Fatigue-critical fillet	Structural	Mask, or thin coating + shot peen	Fatigue penalty is the concern, not dimension
Cable / tendon channel	Ra-critical	Mask, use dry film instead	Roughness increase defeats the purpose

Alloy Selection and Its Effect on Coating Behaviour

Growth behaviour and coating quality both depend on what you’re coating.

6061-T6: The Predictable Baseline

6061 is the reference alloy for anodising, and everything in this article applies to it most directly. Growth ratio is close to the nominal 50/50, coating quality is consistent, colour is uniform, and the finish is repeatable batch to batch.

If your part is going to be hard coated and the alloy choice is open, 6061 is the low-risk option.

7075-T6: Higher Strength, Less Consistent Coating

7075’s zinc content affects coating formation. The practical consequences: colour tends toward a yellowish or inconsistent cast under clear coatings, coating hardness can be somewhat lower than on 6061 at equivalent thickness, and the coating tends to be more brittle — which compounds the fatigue concern on a material often chosen precisely for structural applications.

The dimensional behaviour is broadly similar to 6061, but batch-to-batch consistency is poorer, which makes pre-compensation harder to rely on.

7075 remains the correct choice where its strength is needed. The guidance is to mask precision features rather than compensate, and to be more cautious about hard coating over fatigue-critical regions.

2024 and High-Copper Alloys

2024 and other high-copper alloys are difficult to hard coat well. The copper-rich phases behave differently from the surrounding matrix during anodising, producing coatings that can be patchy, lower in hardness, and less corrosion-resistant than the same process on 6061.

Where 2024 must be hard coated, expect to work more closely with the finisher on process parameters and expect less consistency. In robotics applications, 6061 or 7075 usually serves the requirement without the complication.

Cast Alloys: Why Hard Coating Often Disappoints

Cast aluminium alloys — A380, ADC12, and similar die-casting alloys — contain high silicon content, typically 8-12%. Silicon doesn’t anodise. The result is a coating that forms around silicon particles rather than through them, producing a mottled appearance, reduced hardness, and inconsistent thickness.

This matters for humanoid programmes transitioning shells from machining to die casting. A machined 6061 prototype that hard coats beautifully may become a die-cast A380 production part that hard coats poorly — and the appearance and wear performance both change. If your production part will be cast and your prototype is machined, coat a cast sample before you commit to the finish specification.

How to Specify Hard Anodizing on a Drawing

Most of the failures in this article are drawing failures. Here’s what a complete specification contains.

Before-Coating vs. After-Coating Dimensions

The single most important clarification, and the one most often left ambiguous.

By convention, dimensions on a drawing apply to the finished part — meaning after coating. But this convention is not universally understood, and many engineers dimension parts thinking about the machined state without realising they’ve specified a post-coating requirement.

State it explicitly. A note reading UNLESS OTHERWISE SPECIFIED, DIMENSIONS APPLY AFTER COATING or DIMENSIONS APPLY BEFORE COATING — SEE NOTE 3 FOR COATING ALLOWANCE removes the ambiguity entirely. Which one you choose matters less than that you choose.

Where a specific feature differs from the general rule — a bore dimensioned before coating because it will be masked, for instance — flag it locally at the feature rather than relying on the general note.

Specifying the Coating Standard and Thickness Tolerance

HARD ANODIZE is not a specification. A complete callout includes:

• Standard: MIL-A-8625F Type III (or the applicable regional equivalent)
• Class: Class 1 (undyed) or Class 2 (dyed), with colour if Class 2
• Thickness with tolerance: e.g. 0.0015-0.0025 IN (38-64 μm) — and if you need tighter than the shop default, say so explicitly, because tighter thickness control costs more and won’t happen by default
• Sealing: specify whether the coating is to be sealed, and with what — sealing affects wear and corrosion performance, and dichromate sealing has regulatory implications in some markets

If you’re pre-compensating dimensions, the thickness tolerance stops being an informational detail and becomes a critical dimension. Treat it accordingly.

Calling Out Masked Areas

Masking requires an unambiguous specification. A note saying “mask bearing bores” leaves the boundary location to the finisher’s judgement.

Best practice:

• Mark masked regions on a dedicated view, cross-hatched or shaded
• Dimension the mask boundary explicitly, particularly where a transition zone must clear a functional surface
• State what happens to the masked area — bare, chromate conversion coated, or otherwise protected
• Where a feature will be machined after coating instead of masked, say that rather than leaving it implied

Racking Point Control

Add a note specifying permitted or prohibited racking zones. Something as simple as RACK CONTACT PERMITTED ON SURFACE A ONLY or NO RACK MARKS PERMITTED ON SEALING FACES OR WITHIN 5 MM OF BORE Ø50 H6 prevents contact marks landing somewhere that matters.

Post-Coating Surface Finish Requirements

If a surface has an Ra requirement that must survive coating, state that the requirement applies after coating, and be realistic about what’s achievable. Requesting Ra 0.4 μm after Type III on a surface machined to Ra 0.4 μm is asking for something the process won’t deliver.

Where a fine finish is genuinely required on a coated surface, the route is usually to coat and then lap or hone — which brings back the hard-coating-is-hard-to-machine problem, and usually argues for masking instead.

A Drawing Note Template

A starting point to adapt. Numbers are examples, not recommendations for your part.

SURFACE TREATMENT NOTES

1. HARD ANODIZE PER MIL-A-8625F TYPE III CLASS 1.
   COATING THICKNESS 0.0018-0.0022 IN (46-56 μm).
   SEAL PER [specified method].

2. UNLESS OTHERWISE NOTED, ALL DIMENSIONS APPLY AFTER COATING.

3. MASK AREAS SHOWN CROSS-HATCHED ON VIEW C.
   MASK BOUNDARY TO BE POSITIONED PER DIMENSIONS SHOWN,
   MINIMUM 3 MM CLEAR OF ANY TOLERANCED FEATURE.
   MASKED AREAS: CHROMATE CONVERSION COAT PER [applicable spec].

4. ALL EXTERNAL EDGES TO BE BROKEN 0.8 MM MIN PRIOR TO COATING.

5. RACK CONTACT PERMITTED ON SURFACE D ONLY.
   NO RACK MARKS PERMITTED ON SEALING FACES.

6. THREADED FEATURES TO BE MASKED. TAPPING AFTER COATING
   IS ACCEPTABLE AS AN ALTERNATIVE — NOTIFY BEFORE PROCESSING.

7. COATING THICKNESS TO BE VERIFIED AND REPORTED AT
   LOCATIONS MARKED T1, T2, T3.

Note item 7. Specifying where thickness is measured matters, because thickness varies across a part — a certification reporting nominal thickness measured on a convenient flat face tells you nothing about the deep bore you were worried about.

Managing the Machining-to-Finishing Handoff

Why This Is an Interface Problem, Not a Process Problem

Everything technical in this article is solvable. Growth rates are predictable. Masking works. Compensation works within its limits. The reason parts still fail is organisational.

The machine shop is contracted to produce parts to a drawing. The finisher is contracted to apply a coating to a specification. Between them sits the interaction — coating allowance, mask boundary placement, thickness tolerance, racking constraints, transition zones — and in most supply chains that interaction has no owner.

When the customer manages both suppliers directly, the customer owns it, whether or not they know that. When it goes wrong, both suppliers can demonstrate compliance with their own scope, and the customer absorbs the loss.

What Richconn Does on Robot Component Programmes

We’re a machining company. Surface finishing on our robot component work is performed by vetted finishing partners rather than in our own facility, and we’d rather say that plainly than imply otherwise.

What we do own is the interface. In practice that means:

At design review, we flag coating-dimension conflicts before parts are made. A drawing with an H6 bore and an unqualified hard coat note gets raised as a question, not quoted as-is. This is the cheapest possible point to catch the problem, and it’s where most of the value is.

In specification, we work with you to define the masking strategy, thickness tolerance, mask boundary placement, and racking constraints — and we get those onto the drawing rather than leaving them as verbal agreements that don’t survive a batch change.

Through processing, we manage parts to and from the finisher against that specification, rather than handing them over and hoping.

On return, we inspect coated parts dimensionally against the after-coating requirement — not just the coating certification. Precision features get re-verified after coating, because a coating certificate is not a dimensional report.

The practical effect for you is one supplier accountable for a part that fits, rather than two suppliers each accountable for half of a part that doesn’t.

Questions to Ask Your Supplier

Whether or not you work with us, these questions separate suppliers who manage this interface from those who don’t:

• Who is responsible if a bore is in tolerance after machining and out of tolerance after coating?
• Do you review drawings for coating-dimension conflicts before quoting?
• What coating thickness tolerance can your finisher actually hold, and is it verified on my parts or to a general shop standard?
• Where will thickness be measured, and will I get that data?
• Do you inspect precision features dimensionally after coating, or only review the finishing certification?
• How do you specify and verify mask boundaries?
• If parts come back out of tolerance, what’s the recovery process and who bears the cost?

A supplier who answers these specifically has thought about it. A supplier who answers generally has not.

Conclusion

Hard anodising is an excellent coating. It transforms wear resistance, provides genuine corrosion protection, and on the right surfaces it’s exactly the correct engineering choice. Nothing here argues against using it.

What this article argues is that hard anodising is a dimensional operation, not just a surface operation, and it needs to be designed for with the same discipline as any machining process. A 50 μm coating takes 50 μm out of a bore. That’s not a subtlety — it’s three tolerance bands on a precision fit, and no amount of machining precision recovers it.

The practical summary:

• Diameter change is approximately the coating thickness — the memorable rule
• Threads change by roughly four times the buildup — mask them
• Type II is not dimensionally safe either — 15 μm still breaks an H6 fit at small diameters
• Masking is the default for precision features, not the exception
• Pre-compensation is limited by coating thickness tolerance, which is usually the binding constraint
• The fatigue penalty is real — 30-60% reductions are commonly reported, and it matters on cyclically loaded robot parts
• Most of this is a drawing problem, solved by explicit specification of before/after coating dimensions, thickness tolerance, mask boundaries, and racking zones

And the organisational point underneath all of it: this failure happens in the gap between machining and finishing. Somebody has to own that gap. If your supply chain hasn’t assigned it, it’s yours.

If you’re working through coating specifications on precision robot components — bearing housings, sealed enclosures, sensor mounts, anything with an H7-or-tighter feature and a coating callout — get in touch and we’ll review the drawing before it becomes scrap.