Choosing the fit: clearance, transition, interference

Two parts that touch have a relationship, and that relationship has a name: the fit. A pivot in its hole, a lid on a box, a bearing in its seat — each one is meant for a single thing—to slide, to locate, or to lock in place—and that intent decides how much gap should sit between the two walls. Pick the wrong family, and nothing downstream saves you: a part that should turn seizes up, or a press fit, however carefully drawn, falls apart at the first tug. The fit isn't a number you guess at the end of the design; it's a decision you make at the start, based on what the joint is for.

This article is about that one decision: which fit family your joint calls for, and why the reference the rest of the world uses—the one for metal—is worth reasoning from, even though you won't copy its numbers. The real numbers, the actual tenths of a millimeter you put into the model, come in the next article.

Three families, one question

Every fit falls into one of three families, and you choose between them by answering a single question: does this joint need to move, stay put, or never come apart? Each answer names a family, and each family is fixed by the sign of the gap between shaft and hole.

• Clearance — the shaft is always smaller than the hole, so a gap remains and the parts slide or turn against each other. This is the family of motion: a hinge pivot, a drawer guide, a screw passing through a clearance hole, a lid that drops onto its box. The defining question—does it move?—gets a yes. Inside the family there's a wide range: a pivot that turns often needs a good deal more gap than a lid that only seats, because continuous rotation heats the joint through friction and wears the walls.
• Transition — the gap is essentially zero. The part goes in without forcing and comes out by hand, but while it's seated it stays centered, with no appreciable play. This is the family of temporary location: a dowel aligning two halves of a housing, a lid that must sit concentric yet still open. It neither moves in service nor stays fixed forever.
• Interference (press fit) — the shaft is larger than the hole. You have to force it in, and once it's there the friction between the two walls clamped against each other holds it. This is the family of the permanent: a bearing seat, a knob on its shaft, the boss that receives a heat-set insert (covered in Designing for heat-set inserts).

The boundary between the three is the sign of the gap. Clearance has a positive gap; interference, a negative one; transition lives on that knife-edge at zero where the two walls just graze. And, as you'll see, FDM pushes that whole axis one way, toward the tight side.

The ISO ladder: to reason with, not to copy

Mechanical engineering solved this a century ago with the ISO system of fits: a letter and a number for the hole, another letter and number for the shaft, which together name an exact fit for a given diameter. H7/g6 isn't technical decoration; it's a clearance recipe measured in microns. You'll never type that notation into your CAD model—it doesn't belong there—but the table is still useful for one reason: it puts a name to the relative clearance the trade calls sliding, locating, or press. It gives you a starting point with intent instead of a gap pulled out of thin air.

Read it as a ladder, loosest to tightest, and keep the rung, not the figures:

What's worth taking from this ladder isn't its numbers—in milled aluminum, d9 and p6 are separated by a few microns—but its order. It tells you a sliding fit wants a bit more gap than a locating one, and that between "slides" and "press" there are three distinct steps, not a single jump. That hierarchy of intent is what you carry to the printer. The figures stay behind.

In FDM, the whole ladder shifts tighter

Here's why you can't copy those numbers. A machine shop holds ±0.02–0.05 mm routinely: its error band is comparable to the finest gap in the table, so a sliding g6, when machined, comes out as a sliding g6. An FDM printer doesn't reach that precision. First-layer squish, contraction on cooling, bead width biting into the curves, and material ooze all act together, and your real band is closer to ±0.1–0.2 mm. It's several times wider and, worse, biased: as Holes, pegs and first-layer squish explains, those effects aren't scattered at random—nearly all of them close the hole and fatten the shaft.

The consequence is always the same and always in the same direction. If you print a fit by taking its nominal dimensions from the ISO table, the hole comes out narrower and the shaft thicker than drawn, so the real gap is smaller than the one on the drawing: every fit shifts a rung or two toward the tight side of the ladder. What was a close clearance in metal prints as a rubbing fit; what was a transition prints as light interference, so a part drawn to transition dimensions won't come out by hand: to get transition behavior in FDM you have to start from looser dimensions than the table's k6.

Not all of those shifts weigh the same, nor do they act on the whole joint. The curved-bead bias runs along the entire wall, but first-layer squish is local: it narrows only the mouth of a vertical hole, not its full length. And the sign of the bias depends on orientation: a hole printed standing ends up narrower but cylindrical, while one printed lying down, unsupported inside, sags at the top into an oval — a deformation you can't compensate for by opening up the diameter. The total spread also grows with size—contraction is proportional—so don't apply the same margin to a 3 mm pin as to a 40 mm seat.

You'd think this favors press fits—if everything tightens, the interference comes for free—but the opposite is true: the metal's microns don't survive the process.

That's why the ISO ladder is for reasoning, not measuring. It gives you the family and the relative order; FDM forces you to re-size the gap from scratch, knowing that the process has already pushed your fit toward the tight end and that you'll have to open it up on purpose to compensate.

The family is permanent; the number is local

One asymmetry is worth stating plainly before we close. The family you choose is a lasting decision: it depends only on what the joint is for, and it doesn't change if tomorrow you switch printer, material, or nozzle. A hinge needs clearance on any machine; two housing halves that align but open call for transition regardless of the printer; a bearing seat that must never turn is interference, without exception. You decide it once, and it holds.

The concrete number is the opposite: it's local and perishable. The exact tenths of gap depend on PLA versus PETG, on calibrated flow, on layer height, on how you orient the part on the bed. Change any of those factors and the number moves; the family doesn't.

That's why it pays to separate the two decisions and make them in order. First, function fixes the family—move, locate, or hold—and with it the sign of the gap. Then, and only then, you convert that intent into concrete tenths for your printer. Doing it backwards, starting by guessing a number and hoping the joint works, is how pivots seize and press fits fall loose. Once you're clear on the family of each joint, Real printed clearances turns each of these three intents into a concrete gap, in millimeters, ready for the model.