Guides / Process
How Tight to Tolerance a Brass Part, and Where
September 21, 2026
Brass is more forgiving on tolerance than almost any other machined metal. That makes over-tolerancing easy to get away with technically and expensive to get away with commercially.
What brass holds without trying
On free-cutting brass, on a rigid part, on a well set up lathe:
| Feature | Routine | Achievable with effort |
|---|---|---|
| Turned diameter | plus or minus 0.01 mm | tighter, with a finish pass and temperature control |
| Length along the axis | plus or minus 0.05 mm | plus or minus 0.02 mm |
| Drilled hole | plus or minus 0.05 mm | reamed, plus or minus 0.01 mm |
| Milled feature | plus or minus 0.05 mm | plus or minus 0.02 mm |
| As machined finish | Ra 1.6 to 3.2 | Ra 0.8 or better with a finish pass |
Published typical capability for the material, not a commitment from any shop, and every line assumes a part rigid enough not to deflect.
That first row is why brass is popular. Plus or minus 0.01 mm on a diameter is routine here and a real piece of work in stainless.
Why the numbers get worse
Three things break them, and none is about the alloy.
Slenderness. A long thin part deflects away from the tool. The fix is the machine rather than the tolerance: a Swiss lathe supports the bar at the cutting point and holds numbers a conventional lathe cannot on the same geometry. See Swiss turning brass parts.
Thin walls. Material that was holding the rest of the part in balance gets removed and the remainder relaxes. Brass is not especially bad at this, and no metal is immune.
Workholding. This is the one people underestimate. Brass is soft. A standard vice jaw or chuck jaw that would leave no mark on steel will emboss brass and, worse, squeeze a thin part out of round. Soft jaws, sleeves and lighter clamping all cost setup time. Say on the drawing which faces are allowed to carry jaw marks, and the shop can plan rather than guess.
The most expensive habit on a drawing
Putting a tight tolerance on every dimension.
It feels safe. It is not, because every toleranced dimension is an inspection operation. Somebody measures it, records it and signs for it. On a part carrying forty tight dimensions, inspection can genuinely take longer than machining.
Worse, it hides the important ones. When everything is tight, nothing tells the shop which two features actually matter, so they cannot prioritise the setup around them.
The standard fix is a general tolerance block plus specific callouts.
General tolerances per ISO 2768-m unless stated.
Critical dimensions marked with a box.
Surface finish Ra 3.2 general.
Ra 0.8 on faces A and B (sealing).
That drawing quotes cheaper and produces better parts, because the shop now knows where to spend its attention.
Surface finish, same principle
Ra is the average roughness of the surface and it is cheap to improve in one place and expensive to improve everywhere.
Sensible defaults for brass:
- Ra 3.2 general. What the part comes off at anyway.
- Ra 1.6 where a surface is handled or visible but not critical.
- Ra 0.8 on sealing faces, O-ring grooves and anything that mates under pressure.
- Ra 0.4 or better only where you can name the reason.
Two things quietly raise the requirement without appearing in the finish block.
Plating. A thin conformal layer follows whatever is underneath it, so a plated bright part needs a better machined surface than an as-machined one. See surface finishes and plating for brass.
Sealing. An O-ring reads the surface it sits against. A groove finished at Ra 3.2 leaks in ways that are hard to diagnose.
Edges and burrs are not a finish
Ra says nothing about edges. A part can meet every roughness callout and still carry a burr that fails the assembly.
Specify edges explicitly:
Break all sharp edges 0.2 to 0.4 mm.
Bores free of burrs at all cross hole intersections.
The second line is the one that matters, and it is the one most drawings omit. See the burr inside the bore nobody can see.
Threads sit outside the tolerance block
A thread is not a diameter. It is checked with a gauge, not a micrometer, and the tolerance block does not cover it.
Every thread needs its own callout with the class and, if it seals, the gauge standard. See thread standards on brass fittings.
A drawing checklist
Before you send it out:
- Is there a general tolerance block, or is every dimension individually toleranced?
- Are the genuinely critical dimensions marked so they stand out?
- Is there a general surface finish, with better only where needed?
- Is the edge condition stated?
- Is the internal burr condition stated for every cross hole?
- Does every thread have a class, and is it clear which seal?
- Is it stated which faces may carry workholding marks?
- Do dimensions apply before or after plating?
Most RFQs we see miss four or more of those. Each gap is something a shop has to assume, and shops assume conservatively, which means they assume expensively.
Ready to quote a part? Send the drawing with your tightest tolerance in the tolerance field. If the answer is that only two features are tight, say so in the description. It usually lowers the quotes you get back.
Frequently asked questions
What tolerance does brass hold without extra cost?
On turned features in free-cutting brass, plus or minus 0.01 mm is routine on a well set up lathe, which is tighter than most other metals manage without effort. That is a property of the material rather than a promise from any shop, and it applies to a diameter on a rigid part, not to a thin wall or a long slender feature.
Why does a tolerance on every dimension cost so much?
Because every toleranced dimension is an inspection operation. Someone measures it, records it and decides. On a part with forty dimensions all carrying tight tolerances, the inspection can take longer than the machining. Most of those dimensions are not functional and a general tolerance block would cover them.
Does brass move after machining?
Less than most metals, because cutting forces and cutting temperatures are low. Where it does move is on thin walls that were holding residual stress, and on parts clamped hard enough to distort. Brass is soft enough to mark and deform in a vice, so workholding is often a bigger accuracy problem than the cutting is.
How do I specify surface finish sensibly?
Give a general finish for the part and call out the two or three faces that actually need better. Ra 3.2 as a general finish and Ra 0.8 on a sealing face is a normal and inexpensive combination. A blanket Ra 0.8 over the whole part multiplies finishing passes and inspection for no functional gain.
Need a quote for this part?
Send the drawing. We match you with up to 3 brass shops and reply within 24 business hours.
Get matched with a brass machining shop