FDM Tolerances: How to Design Parts That Fit Every Time

Most failed multi-part prints are not bad models. They are bad clearances. If your lid binds, your pin falls out, or your handle refuses to latch, the problem is almost always the gap you left between mating surfaces, not the geometry itself. A new guide from Sovol walks through the numbers, the slicer settings, and the material-specific quirks that separate parts that fit first try from parts that end up in the scrap bin.

Clearance Is Not Optional

FDM printers have real tolerances. A well-calibrated desktop machine holds roughly plus or minus 0.2 to 0.5 mm on most features, with smaller features drifting further relative to their size. Internal holes are the biggest problem: the slicer approximates a circle with straight line segments, and those segments sit inside the true curve. The result is a hole that prints 0.1 to 0.4 mm undersize on diameter, with the error getting worse on small features.

The fix is not to shave the pin. It is to size the hole larger than the pin by the amount your joint actually needs. That extra space is clearance, and it is the single most important number in any functional FDM design.

Starting Points for Press, Sliding, and Snap Fits

Not every joint needs the same gap. A press fit that holds bearings or heat-set bosses wants firm interference, usually 0.05 to 0.15 mm per side. A sliding lid or drawer guide needs room to move, so start at 0.2 to 0.3 mm per side. Snap-fit clips and battery covers rely on deflection, not just gap, so plan at least 0.3 mm plus the geometry needed for the cantilever arm to bend.

Material matters more than most people account for. PLA is the most stable, with low shrinkage and predictable dimensions. PETG shrinks a little more and can string into tight gaps, so add 0.05 mm per side compared to PLA. ASA is the most sensitive, with significant thermal shrinkage that can close a carefully designed gap if the printer is not enclosed and cooled consistently.

Six Things That Change the Fit Without You Touching the Model

Nozzle diameter changes how precise thin walls and small holes can be. Smaller nozzles hold tighter clearances; larger nozzles make fine gaps blurrier. Print orientation matters because the XY plane is dimensionally strongest, while layer boundaries create weakness and slight steps. Elephant's foot, the bulge created by first-layer squish, can turn a 0.2 mm clearance into a press fit at the base of a part. Over-extrusion thickens every wall and closes gaps, while under-extrusion leaves holes oversized and pins loose. Layer height controls vertical fidelity, so a 0.25 mm gap at 0.2 mm layers effectively becomes 0.2 mm or 0.4 mm depending on how the slicer rounds.

Slicer settings can fix systemic problems without editing the CAD. XY compensation, also called horizontal expansion, nudges all outlines in or out in the XY plane. Elephant's foot compensation reduces the bulge on the first few layers. Both are worth testing before you reprint the full model.

Print a Tolerance Test Before the Final Part

The only clearance number that matters is the one your specific printer and material actually produce. A tolerance test is a small coupon with graduated holes and matching pegs, stepping from 0.1 mm to 0.5 mm in 0.05 or 0.1 mm increments. Print it with the same settings you will use for the final part, then record which gap feels like a press fit, which slides smoothly, and which is loose. Those three numbers become your baseline for every functional design you run on that machine with that filament.

Troubleshooting Tight and Loose Joints

If a part is too tight, check for over-extrusion first. Calibrate your flow rate, enable elephant's foot compensation, and add a small chamfer to the hole entry so the parts guide themselves into place. If a part is too loose, you may have accidentally used per-side clearance as total clearance, doubling the gap. Reduce the modeled clearance by 0.1 mm steps and retest. Warped mating faces usually come from poor bed adhesion, drafts, or wrong bed temperature, not the model itself.

The practical takeaway is simple: never trust a default clearance number. Print a ten-minute test coupon, read your actual baselines, and apply them to the model. That small step is what separates a part that binds on the bench from one that assembles first try.