Tolerances and Clearances: Making Printed Parts Actually Fit Together

The most common disappointment in 3D printing is an assembly that doesn't go together. Two parts, both modelled correctly, both printed correctly, and the shaft won't enter the hole.

Nothing has gone wrong. The parts were designed as though they would be manufactured to their nominal dimensions, and FDM doesn't work that way. This article explains what the process actually delivers, what clearances to design in, and how to get an assembly right on the first print rather than the third.

What tolerance you actually get

We hold ±0.20 mm, or ±0.3% of the nominal dimension, whichever is greater.

Read that carefully, because the percentage term surprises people on large parts:

• On a 20 mm feature: ±0.20 mm
• On a 50 mm feature: ±0.20 mm (0.3% would be 0.15 mm, so the 0.20 floor applies)
• On a 100 mm feature: ±0.30 mm
• On a 200 mm span: ±0.60 mm

That last figure is the one that catches assembly designs. If you've two mounting holes 200 mm apart and they must align with holes in a metal chassis, ±0.6 mm of positional error is a real problem. Design a slot instead of a round hole for one of them and the problem disappears.

This is a good tolerance for a layered thermoplastic process. It isn't machining. If you need ±0.05 mm, print the part and then machine or ream the critical feature.

Why holes come out undersized, always

This is the single most reliable dimensional behaviour in FDM, and once you know it you can design around it permanently.

Ask for a 5 mm hole. Measure it. You'll get somewhere between 4.7 and 4.9 mm.

The cause is geometry, not calibration. The nozzle traces the circumference of the hole while extruding a bead about 0.4 mm wide. On the inside of a curve, the material on the inner edge of that bead is being laid on a tighter radius than the nozzle centre-line, so it squeezes inward slightly. The tighter the radius, the more pronounced the effect.

Consequences worth memorising:

• Small holes lose proportionally more. A 3 mm hole might come out at 2.7 mm: a 10% error. A 20 mm hole might come out at 19.8 mm: a 1% error.
• The effect is consistent, so it can be compensated reliably.
• It applies to internal features generally, not just round holes. Slots, pockets and internal corners all come out slightly tight.

The fix: oversize holes by 0.2-0.3 mm in CAD. If an M5 bolt must pass through, model the hole at 5.4 mm rather than 5.0 mm. If a bearing must press in, this isn't the approach, see the press-fit section below.

For anything that must be accurate and round, the better answer is to print an undersized pilot and drill it out. A drilled hole in a printed part is round, straight and on-size in a way a printed one never quite is, and drilling plastic takes seconds.

Clearances for each kind of fit

These are total clearances across the joint, the difference between the hole and the shaft, or the pocket and the part. Split them between the two components, or apply all of it to one; the total is what matters.

Fit	Total clearance	Behaviour
Free-running	0.4-0.5 mm	Rotates or slides freely, visible play
Sliding	0.3 mm	Moves smoothly with light resistance
Snug	0.15-0.2 mm	Assembles by hand, no play, no force
Press fit	0.05-0.1 mm interference	Needs force, stays put
Print-in-place	0.25-0.35 mm	Separate parts printed as one assembly

A few notes on using these.

Start looser than you think. A joint that's slightly loose is usable; a joint that's too tight isn't, and forcing it will crack the part. If you're guessing, guess loose.

Print-in-place needs more clearance than you'd expect because there's no way to clean up between the surfaces afterwards. 0.25 mm is about the minimum that reliably comes free; below 0.2 mm the parts fuse and you've printed a solid lump.

Press fits in plastic aren't press fits in metal. Plastic creeps, under sustained stress it slowly deforms, and a press fit that was tight on day one may be loose after six months. For anything that must stay tight indefinitely, use a mechanical retention feature or an adhesive rather than relying on interference alone.

Test coupons: the trick worth adopting

If a fit is critical and you get one attempt, don't print the whole assembly and hope.

Print a small test coupon of just the mating feature. A 20 mm cube with the hole in it, or a short section of the rail, costs almost nothing and takes minutes. Try the fit, adjust the CAD, and only then commit to the full part.

Better still, print a coupon with several clearances at once. The same hole at 0.1, 0.2, 0.3, 0.4 and 0.5 mm clearance, labelled. One small print tells you exactly what your specific mating part needs, and you can keep the coupon as a reference for future designs.

This is standard practice in shops that do a lot of assembly work, and it converts fit from guesswork into measurement. Given our ₹350 order minimum, a coupon costs less than a single failed full-size print.

The elephant's foot problem

The first layer of a print is deliberately squashed into the build plate to make it stick. That squash spreads the plastic wider than nominal, and the effect persists through the first two or three layers, producing a slight flare at the base, usually 0.1-0.2 mm.

For most parts, irrelevant. For an assembly, it matters in two specific cases:

A part that must sit flat in a pocket. The flared base catches on the pocket edge and the part won't seat.

A press fit that starts at the bottom face. The first few millimetres are tighter than the rest, so the part jams on entry.

The fix is a 0.5 mm chamfer on the bottom edge. It gives the spread somewhere to go, and it also makes the part easier to remove from the plate. This is close to free and it solves the problem completely.

Orientation affects accuracy

Dimensions in the XY plane, across the build plate, are more accurate than dimensions in Z, because XY is controlled by precise motion of the toolhead while Z is controlled by layer stacking.

Practical consequences:

• A hole printed with its axis vertical (so it's drawn as a circle in each layer) is rounder and more accurate than the same hole printed horizontally, where it's built from stacked layers approximating a curve.
• A horizontal hole will also have a flat spot at the top where the unsupported layers droop. Design it as a teardrop rather than a circle if it must be printed that way.
• Overall height is generally the most accurate dimension of all, because it's a count of layers.

If a specific hole must be accurate, tell us which one and we'll orient the part so it prints vertically. That's free and it makes a measurable difference.

Designing assemblies that tolerate error

The most robust approach isn't to chase tight tolerances at all, but to design assemblies that don't need them.

Use slots instead of holes for one of any pair of mounting points. A slot absorbs positional error entirely. This single change fixes most alignment problems on larger parts.

Design self-aligning features. A tapered lead-in, a chamfered entry, a conical boss that guides itself into a matching recess. These convert a precise positioning problem into an approximate one.

Locate from one datum. If every feature is dimensioned from the same face, errors don't accumulate. If features are chained, this one 20 mm from that one, which is 30 mm from the next, the tolerances stack and the far end can be well out.

Use fasteners with built-in adjustment. Clearance holes with washers absorb far more error than a tight bolt hole.

Split the tolerance across both parts. If a shaft must fit a hole, take material off the shaft and add it to the hole rather than putting all the clearance in one place. Each part then deviates less from nominal.

When printed accuracy isn't enough

Be realistic about the limits. If your design needs any of the following, plan a secondary operation:

• Better than ±0.1 mm on a specific feature → drill, ream or mill it after printing
• A genuine bearing seat → print undersized and ream, or design for a pressed-in metal sleeve
• A reliable, repeatedly-used thread → heat-set insert, not a printed thread
• A precision sliding surface → consider a metal rail or shaft with printed brackets holding it

This isn't a failure of the process. It's how printed parts are used in practice. The printed component provides the complex geometry cheaply; a metal insert or a machining operation provides the precision at the two or three places that need it. That combination is usually far cheaper than machining the whole thing.

A pre-flight checklist for assemblies

1. Have you added clearance, or modelled to nominal? Nominal won't fit.
2. Are through-holes oversized by 0.25 mm?
3. Do bottom edges have a 0.5 mm chamfer?
4. Is one hole in each mounting pair a slot?
5. Are features dimensioned from a common datum rather than chained?
6. Over a 200 mm span, can the design tolerate ±0.6 mm?
7. Is there a lead-in chamfer on anything that has to be inserted?
8. Have you printed a test coupon for the critical fit?

Tell us what has to fit

When you send a model, the most useful thing you can say is which dimension is critical and what it mates with. "The 12 mm bore must take a bearing" tells us to orient that hole vertically, flag it in review, and consider whether to print it undersized for reaming.

Without that, we optimise for surface finish and support minimisation, and we make reasonable assumptions. But we can't prioritise a dimension we don't know matters.

Design-for-additive review is free with every quote, and fit problems are the single most common thing it catches. Sending the mating part's dimensions along with the model costs you one sentence and saves a reprint.