What Is PM Tolerance Capability?

Here’s the thing about tolerance in powder metallurgy: the die makes the part. Most dimensions never meet a cutting tool. So the question “how tight can PM hold?” isn’t a machine-precision question the way it is in CNC work — it’s a process-control question, answered in a powder blending room and a sintering furnace, not at a machining center.

We’ve quoted a lot of drawings over the years, and the same pattern keeps showing up. Engineers who understand where PM tolerance actually comes from spec their parts so the process holds dimensions naturally, and they pay almost nothing for accuracy. Engineers who don’t — usually by copying tolerance callouts from a machined part drawing — end up paying for sizing and CNC on dimensions that never needed either.

This article is the explanation we wish every RFQ came with. Where the variation comes from, what each process step holds, and how to write a PM drawing that doesn’t cost more than it should.

The PM Tolerance Chart

Every step after compaction gives some of the tolerance back; sizing is the step that claws it back. Here’s what each stage holds in practice:

A few things the table doesn’t say on its own.

Dimensions created by the die move less than dimensions created by the powder. That sounds obvious once said, but it changes how you should read the table. An outer diameter or a die-formed bore is a copy of a precision tool. A height dimension is a copy of however much powder happened to fill the cavity — so heights are always the loosest numbers on the chart, and no amount of press tuning changes that fundamentally.

The other quiet observation: sizing doesn’t help everything equally. Re-pressing a bore works beautifully because the sizing punch registers off a die-formed surface. Re-pressing a height that depends on fill weight helps, but less dramatically.

Where the Variation Actually Comes From

Four things stack up to determine your tolerance, and they’re not equal contributors.

The powder starts it. Particle size distribution, apparent density, flow rate — these decide how evenly a die cavity fills. A powder lot that flows slightly differently fills thin sections slightly differently, and that density difference reappears after sintering as dimensional scatter. None of this shows up on a powder datasheet. It shows up in whether your supplier specifies incoming powder tightly and verifies every lot, or buys on price and hopes.

Compaction comes next. Fill height control and press repeatability set how consistent the green parts are. Two points worth knowing. First, higher and more uniform pressed density means more predictable sintering — a part pressed uniformly to 7.0 g/cm³ shrinks uniformly, while a part with a 0.15 g/cm³ spread across its section distorts. Second, tool wear sounds like a problem but mostly isn’t: it drifts slowly in one direction, and SPC catches drift long before it catches random scatter.

Then sintering — and this is the big one. The part shrinks (or grows) 0.2–0.5% linearly depending on material and green density, and it does this on a belt, inside a furnace, surrounded by hundreds of other parts. Any temperature gradient across the furnace, any atmosphere non-uniformity, any difference in where the part sits on the belt becomes dimensional scatter across the load. This is why two suppliers quoting the same powder and the same drawing can hold noticeably different tolerances: the difference lives in the furnace, and you can’t see a furnace on a website.

The shrinkage itself isn’t the enemy, by the way. It’s remarkably repeatable per material, so it gets compensated in the die — the same trick MIM plays with its far larger 15–20% shrinkage. The enemy is variation in shrinkage, and that’s a furnace-control question. Suppliers who have run your material for years have the shrinkage data; suppliers quoting your first run are about to find out.

Put together, these four contributors explain why powder metallurgy dimensional accuracy is a process-control conversation, not a machine-precision one. The as-sintered number you actually get in production is the sum of powder lot, press repeatability, furnace uniformity, and the discipline of the supplier running your part.

Sizing and CNC buy back what sintering cost. Re-pressing the sintered part in a second die gets selected dimensions to ±0.025–0.05 mm. It costs a die, a press operation, and lead time — which is why the rule is to apply it only where the part’s function requires it. A bearing bore, yes. A bracket width, no.

Tolerancing a PM Drawing Without Overpaying

After enough DFM reviews, we’ve settled on a short list we’d hand to any customer:

1. Start every dimension at as-sintered — ±0.1 mm typical, ±0.05 mm from a controlled process. Tighten selectively. Never globally.
2. Put the tight tolerances on die-formed features. A die-controlled bore sized to ±0.025 mm is cheap. A height dimension held that tight is not, and never will be.
3. Thin walls and long unsupported sections distort in sintering no matter who runs the press. If a dimension on such a feature must be tight, plan sizing or CNC from day one instead of discovering it after tooling.
4. Use die-formed datums. Tolerances chained across features that shrink independently never hold the way the drawing implies. Reference criticals to the die-formed OD or bore.
5. Anything tighter than ±0.025 mm is a machined feature. PM competes by not machining. If a dimension needs less than that, your supplier should price the machining openly from the first quote — some do, some hide it in the unit price.
6. Send the drawing before the tooling decision. A DFM review marks each tolerance as die-limited, sinter-limited, or needing sizing/CNC. Doing this after the die is cut is called re-cutting the die.

How PM Tolerance Compares to Machining and Casting

The economics are the interesting part. Machining buys accuracy on every dimension whether you need it or not; PM gives you most dimensions nearly free and charges for the exceptions. A PM gear with a sized bore and as-sintered teeth at 50,000 pieces costs a fraction of a fully machined one. The same part with three machined faces starts giving that advantage back, dimension by dimension. So the real DFM question is never “can PM hold this?” — it’s “which of these dimensions actually needs a cutting tool?”