Treat +/- 0.005 mm as a feature-level precision band, not a blanket CNC default. The cases when +/- 0.005 mm CNC tolerance is realistic are short, rigid metal features: a finish-bored hole with L/D of 3 or less, a turned or ground diameter with good support, and a small datum face cut with a short end mill. 6061-T6, 7075-T6, 304/316 stainless, and 4140 can hold it when the shop inspects with a CMM or air gauge and the drawing follows ASME Y14.5-2018. ISO 2768-1 (1989) fine class is still about ±0.05 mm to ±0.1 mm on small sizes, so 5 µm must be called out. Keep large milled faces, walls under 1.5 mm, plastics, and Type II anodized sizes (MIL-A-8625F, 1993; 5-25 µm) at ±0.02 mm or looser unless you add grind or jig-bore.
This guide is for a US or EU engineer sending a drawing to a China factory and deciding which sizes deserve 5 µm. You are not making the whole part “high precision.” You are protecting fits, datums, and the features you will inspect.
If every face is ±0.005 mm, the quote assumes extra setups, extra CMM time, and extra scrap risk. If only the right faces carry that band, custom CNC machining services can plan a finish pass and a gauge instead of guessing.
Key takeaways
- ±0.005 mm is a 10 µm total band, about 10 times tighter than ISO 2768-1 (1989) fine class on small linear sizes. It is a special callout, not a title-block default.
- Hold it on short, closed, rigid metal features you can probe or gauge at a stated temperature. Do not hold it on large as-milled faces, thin walls, plastics, or coated sizes.
- Write the inspection method next to the size. A CMM, air gauge, or micrometer is part of the tolerance. A number with no gauge is a wish.
- Temperature and coating eat the band faster than most CAD defaults. Aluminum moves about 23 µm per meter per °C. Type II anodize builds 5-25 µm.
What ±0.005 mm means against ISO 2768
±0.005 mm is a 10 µm total band. On a 10 mm bore that is a fit class, not a general mill finish.
ISO 2768-1 (1989) still covers unmarked linear sizes. Fine (f) class allows ±0.05 mm from 0.5 mm to 6 mm and ±0.1 mm from 6 mm to 30 mm. Medium (m), a common shop default, is ±0.1 mm from 0.5 mm to 6 mm and ±0.2 mm from 6 mm to 30 mm. A title block that only says ISO 2768-m never implies 5 µm. Fine class on a 20 mm length is still ±0.1 mm, twenty times wider than ±0.005 mm.
Call the 5 µm features with numbers, datums, and a measurement note per ASME Y14.5-2018. Leave the rest on the general class. That split keeps function and quote time under control.
A 10 µm band also has a temperature story. Typical published expansion is about 23 µm/m/°C for aluminum, about 17 µm/m/°C for 304 stainless, and about 12 µm/m/°C for carbon steel. On a 100 mm 6061 length, 1 °C of shop-to-CMM difference is about 2.3 µm. That is why 5 µm belongs on short features, and why inspection notes often say 20 °C. H7 on a 10-18 mm hole is +18/0 µm, already looser than ±0.005 mm. If the assembly is a handbook fit, write the class. Do not convert H7 into 5 µm “to be safe.”
When +/- 0.005 mm CNC tolerance is realistic by feature
When +/- 0.005 mm CNC tolerance is realistic, the feature is small, round or well supported, and cut with a short, stiff tool.
Bores Ø6-20 mm with L/D ≤ 3, finish-bored or reamed, are the usual yes. A Ø10 mm bore 25 mm deep is L/D 2.5 and is a candidate after a rough pass and a single-point finish bore. A Ø8 mm bore 40 mm deep is L/D 5. The extra stick-out bends the bar. Size and roundness then wander more than 5 µm unless you hone or jig-bore.
Turned diameters with L/D ≤ 4 and a tailstock or short stick-out are often yes after a finish pass. A Ø16 mm journal 40 mm long is L/D 2.5: finish turn can work. The same Ø16 mm at 80 mm long is L/D 5: leave 0.1-0.2 mm grind stock if 5 µm is real. Small datum pads a CMM can hit in one setup are often yes if the pad is 15-25 mm across, not an 80 mm open face. A 20 mm pad cut with a 12 mm end mill at 2-3× diameter stick-out is a different problem from facing 120 mm of plate with a fly cutter.
Open milled faces larger than about 50 mm, deep pockets that need a long end mill, thin walls, and any size measured after Type II anodize are the usual no. Cutter deflection, heat, and coating (5-25 µm) are larger than half of a 10 µm band. A 63 mm face mill on a 80 mm cover can leave a 0.01-0.03 mm dish without a programming error. If the mill cannot see the feature with a short tool, name grind, hone, or jig-bore as an operation, not as hope. Threads and chamfers belong on their class or ISO 2768, not on 5 µm.
CNC milling can hold tight local features on a large plate while overall length stays on ISO 2768. CNC turning can hold a journal to 5 µm while shaft length stays looser. Axis count does not create 5 µm. A 5-axis program with a 60 mm stick-out still bends.
When +/- 0.005 mm CNC tolerance is realistic by material and size
When +/- 0.005 mm CNC tolerance is realistic by material, you are cutting a stable metal in a size the shop can fixture without springing it.
6061-T6 and 7075-T6 hold size well on small bosses and holes. They move on thin plates after heavy pocketing. A 6 mm 6061 plate with 4 mm pockets left as a 1.5 mm floor will walk after the roughing heat leaves. You can still hold a Ø8 mm bore to 5 µm in a thick boss on that same plate if the boss is not in the skinny floor. 7075-T6 is stronger and often holds a small bore, but residual stress in plate can shift a thin web 0.02-0.05 mm overnight. That shift is a distortion problem, not a “the mill is inaccurate” problem.
304/316 can hold 5 µm on a short turned diameter if you plan a finish pass. Stainless work-hardens. A worn insert leaves a tapered journal. The fix is a light final pass (0.05-0.10 mm radial) with a fresh edge, not a tighter number on the PDF. 4140 pre-hard is a good candidate for ground journals. As-turned 4140 can hold 5 µm on a short, supported diameter. Long shafts still want grind stock.
POM, nylon, and most unfilled plastics are a no in production air. Moisture and temperature shift the part more than 5 µm. POM’s expansion is several times that of aluminum. A 50 mm POM length can move well beyond 10 µm between a 18 °C stockroom and a 24 °C CMM room. If a plastic bore must locate a pin, use a metal insert or open the band to ±0.05 mm and design the fit for that.
Stock size matters as much as alloy. A Ø20 × 15 mm 304 bushing is a 5 µm candidate. A 200 × 150 × 12 mm 6061 plate with one “precision” overall length is not. Type II anodize per MIL-A-8625F (1993) is not a size process. If you need 5 µm after heat treat on 4140, grind after heat treat.
Features that should stay looser on the same drawing
Most faces on a housing, bracket, or plate should stay at ±0.05 mm to ±0.2 mm, or on ISO 2768-m.
Keep looser: overall length of a mill-cut plate, pocket floors that only clear a cover, thread minor diameters covered by a tap spec, cosmetic chamfers, and any size that will be coated. Type II anodize per MIL-A-8625F (1993) builds about 5-25 µm. That coating can consume the entire ±0.005 mm band by itself. A “black anodize, dimensions apply after coat, Ø10.000 ±0.005” note is how incoming inspection and the dye tank fight each other.
| Feature | ±0.005 mm? | Better default | Why |
|---|---|---|---|
| Bore Ø6-20 mm, L/D ≤ 3, metal, before coat | Yes, finish bore + gauge | Keep ±0.005 | Short tool, closed round feature |
| Turned journal, L/D ≤ 4, metal | Often, finish turn or grind | ±0.005 after grind stock | Support and roundness are controllable |
| Milled face > 50 mm, as-milled | No | ±0.02 to ISO 2768-m | Heat, tool wear, fixture |
| Wall < 1.5 mm | No | ±0.05 or profile 0.1 | Chatter and springback |
| Size after Type II anodize | Rarely | Dimension before coat, or ±0.02 coated | Coating 5-25 µm |
| POM / nylon | No | ±0.05 or wider | Moisture and creep |
| Overall plate length 80-200 mm | No | ISO 2768-m | Thermal and fixture stack |
| Tap drill / thread minor | No | Thread class (e.g. 6H) | The tap owns that size |
Bottom line: put ±0.005 mm only on the fits and datums you will inspect. Leave the rest on a general class.
A useful split on a 6061 housing is three bands: ISO 2768-m on unmarked sizes, ±0.02 mm or ±0.05 mm on functional but non-fit faces, and ±0.005 mm on two or three locating bores. True position 0.05 mm at MMC on a hole pattern often assembles better than four independent 5 µm hole sizes with no datums. Send the marked PDF to YXT CNC’s contact form if you want a process note on which 5 µm callouts should stay.
RFQ notes that stop a false yes
A China factory can say “we can do ±0.005 mm” and still mean one hole on a sample, not every face on a 200-piece lot. Freeze the inspection story in the RFQ so the yes is about production, not a hero first article.
- Highlight every ±0.005 mm size on the PDF and name the gauge (CMM, air gauge, micrometer) plus 20 °C if the feature is metal. Done when a programmer can count the precision features without hunting, and each one has a method.
- Check L/D, wall thickness, and tool reach in the STEP. Done when precision bores are L/D ≤ 3, nearby walls are ≥ 1.5 mm, and tool stick-out is under about 4× diameter, or you have named a grind or hone op.
- Separate coating from size in a note. Done when the drawing says “dimensions apply before anodize” or gives a coated limit that includes 5-25 µm of Type II build.
- State quantity and inspection scope (FAI only vs a sampling plan, and which balloons are 100%). Done when the RFQ does not imply 100% CMM on unmarked faces.
- State whether size is in-process or after all secondary ops (bead blast, anodize, heat treat). Done when the quote cannot hide a 8 µm blast texture inside a 10 µm band.
If a feature fails that check, loosen it or add an operation. Do not keep the 5 µm number as decoration. Ask the shop to confirm the gauge, not to “confirm capability.” For bores, an air gauge at 20 °C is often more repeatable than a CMM stylus in a deep hole. Use the CMM for datum-related true position, not as a default micrometer.
Worked RFQ: mixed 5 µm and ISO 2768 on one housing
Here is a drawing you can copy as a pattern. Part P-18420, 6061-T6 manifold cover, 120 × 80 × 22 mm, qty 5 / 25 / 100, Type II black 8-12 µm per MIL-A-8625F (1993), dimensions apply before coat, mask 2× Ø12 locating bores.
Title block: ISO 2768-m for unmarked sizes. Datum A is the seal face, datum B is one Ø12 bore, datum C is the second Ø12 bore. Only two sizes are ±0.005 mm: both Ø12.000 locating bores, 18 mm deep (L/D 1.5). Inspection: air gauge at 20 °C on those two bores, CMM true position 0.05 mm to A-B-C on the M4 pattern, FAI ballooned PDF, sampling on later lots.
Everything else is wider. Overall 120 mm stays on ISO 2768-m (medium class on 30-120 mm is ±0.3 mm). Pocket floors that clear a gasket are ±0.1 mm. Walls 2.0 mm thick are ±0.05 mm. M4×0.7-6H threads use the class, not a 5 µm minor. The 80 × 50 mm outer face is ±0.05 mm, not 5 µm, because it is an open mill face.
| Balloon | Feature | Callout | Gauge | Pass / fail rule |
|---|---|---|---|---|
| 1-2 | Ø12 locating bores, 18 mm deep | Ø12.000 ±0.005, before coat, mask | Air gauge, 20 °C | Both bores in band on FAI and on sampled lots |
| 3 | True position of M4 pattern | TP 0.05 to A-B-C | CMM | FAI plus agreed sample |
| 4 | Overall 120 mm | ISO 2768-m | Caliper or CMM | Not a reject unless it misses the class |
| 5 | Pocket floor | ±0.10 | Height gauge | Clearance only |
| 6 | Type II 8-12 µm, black | Before-coat sizes | Coating thickness on a coupon or non-fit face | Fits measured before dye |
That package is quotable. Send STEP AP242 plus PDF Rev C so the estimator can price two finish bores, masking plugs, and a short FAI. The false version puts ±0.005 mm in the title block, anodize after size with no mask, and “CMM 100%.” That version fails incoming on a 120 mm length that should never have carried the band.
FAQ
Can a China CNC shop hold +/- 0.005 mm on every dimension?
No. Hold ±0.005 mm on selected metal features with a finish pass and a matching gauge. Unmarked faces stay on ISO 2768-1 (1989). Shops that answer “yes” to a blanket 5 µm title block are usually talking about one bored hole on a sample, not 200 pieces of every face. Put the 5 µm on the fits you will reject, name the gauge, and leave the rest on a general class.
Is +/- 0.005 mm the same as ISO 2768 fine?
No. ISO 2768-1 (1989) fine class is about ±0.05 mm to ±0.1 mm on small sizes. ±0.005 mm is a separate callout. Fine class on 6-30 mm is ±0.1 mm, twenty times the 5 µm bilateral band. Medium class, the usual shop default, is wider still. Write ISO 2768-m or ISO 2768-f in the title block for unmarked sizes, then number the precision features. Do not assume a “fine” note upgrades the whole part to gauge-maker work.
Can I hold +/- 0.005 mm after Type II anodize?
Usually no on the coated size. Type II per MIL-A-8625F (1993) is 5-25 µm, as thick as the whole band. Dimension before coat. If a locating bore must stay 5 µm after dye, mask it, or grind/hone after coat and accept that as a named operation. Color and rack marks do not make the coating uniform to 2 µm. A note that says “anodize black” with no before/after rule will be quoted as as-machined size plus a tank, and incoming will measure a different hole. Put mask, thickness range, and the size rule on the PDF.
Does 6061-T6 hold 5 µm better than stainless?
On small rigid features, 6061-T6 is often easier. 304/316 can still hold 5 µm on a short turned diameter with a finish pass. Aluminum cuts with lower force, so a short end mill or boring bar deflects less. Stainless work-hardens and wants a sharp edge and a planned final pass. Neither alloy holds 5 µm on a 2 mm wall or a 100 mm open face. Compare the feature and the tool, not the alloy slogan. 7075-T6 is also a candidate on small bores, with more plate-movement risk after heavy pockets.
What inspection method should I specify for +/- 0.005 mm?
Name the method: air gauge or bore gauge for holes, micrometer for journals, CMM for datum-related true position. A CMM is not automatically better. A 1 mm stylus in a deep bore adds cosine and bending error you do not have with an air plug. State 20 °C for metal fits. State whether FAI is 100% of balloons and whether production lots are sampled. If you will reject on a printout the shop never quoted, the first article fight is already scheduled. Write the gauge next to the 5 µm size so the programmer and the inspector share one story.
Should plastic CNC parts use +/- 0.005 mm?
No for production atmosphere. POM and nylon shift with moisture and temperature more than 5 µm. Expansion coefficients for acetal are several times aluminum, so a 40-50 mm plastic length can leave a 10 µm band between rooms. If a plastic part must locate metal hardware, use a machined insert, a pressed bushing, or open the band to ±0.05 mm and design the fit for plastic. You can still mill POM to a nice cosmetic face. That is not the same as a production 5 µm bore in humid shipping air.
How many 5 µm callouts should one drawing carry?
As few as the assembly needs, often one to four features. Two locating bores and one journal is a normal precision set. Twenty faces at ±0.005 mm is a title-block accident. Each 5 µm size needs a finish method and a gauge, which adds cycle time and FAI time even when the mill could hit the number on a good day. If a size would not scrap the lot, it does not belong in the 5 µm club.
Does 5-axis machining make +/- 0.005 mm realistic on large faces?
No. Extra axes change access and datum transfers. They do not cancel cutter deflection, heat, or coating. A 5-axis finish pass with a long tool on a 80 mm open wall is still an as-milled face. Hold 5 µm only on short rigid features you can gauge. If a large face must be 5 µm flat, name grind, not a spindle-count upgrade.
Decide when +/- 0.005 mm CNC tolerance is realistic by feature, not by slogan: short rigid metal, short tools, a named gauge, and a before-coat rule. Loosen large faces, thin walls, plastics, and coated sizes. Send the marked STEP and PDF through YXT CNC contact if you want a DFM note on which 5 µm callouts should stay.
