Understanding Tolerances: Why 0.1mm Matters in Precision Engineering

There's a story every machinist knows, or learns the hard way: you spend six hours turning a shaft on the lathe, check the diameter with your Vernier caliper, it reads exactly 25mm, you press it into the bore — and it either falls straight through or won't go in at all. The drawing said 25mm. The part measures 25mm. So what went wrong?

The answer is tolerances, and understanding them is the difference between an engineer who produces assemblies and one who produces expensive scrap.

What a Tolerance Actually Is

A tolerance is not an admission of failure. It's a deliberately specified range of acceptable variation — the engineering acknowledgment that no manufacturing process produces a theoretically perfect dimension every single time. When a drawing specifies a shaft diameter as 25.000 ±0.025mm, it's saying the finished part may measure anywhere between 24.975mm and 25.025mm and will still function as intended. That 0.05mm total band is the tolerance zone.

Sounds generous, right? Consider that a human hair is roughly 70 microns (0.07mm) in diameter. A 0.025mm deviation is less than half a hair's width. Hold that thought while you run your fingers across a surface that feels perfectly smooth — it's likely undulating by several microns across every millimeter. Manufacturing at this scale means working at the edge of what you can perceive with bare hands.

The ISO System of Limits and Fits

Rather than leaving every engineer to invent their own tolerance conventions, the ISO 286 standard (and its ANSI equivalent, B4.1) established a universal coding system for fits between mating parts — specifically shafts and holes, the foundational pairing in mechanical assemblies.

The system uses a letter-number combination. The letter denotes the fundamental deviation (how far the tolerance zone sits from the nominal size), and the number denotes the grade of tolerance (how wide that zone is). IT grades run from IT01 — reserved for gauge blocks and reference standards — through IT18, which is rough casting territory. Most precision mechanical work lives in IT5 through IT9.

So when a drawing specifies a fit as H7/f6, it's communicating a complete functional relationship: H7 is the hole tolerance (H means the fundamental deviation starts at zero, moving upward from nominal; 7 is a medium-precision grade), and f6 is the shaft tolerance (f means the shaft runs smaller than nominal by a calculated offset; 6 is a tighter grade). Together they define a running/sliding fit — parts that can move relative to each other with a reliable clearance of 0.020mm to 0.053mm for a 25mm nominal size.

That fit would be suitable for a pump shaft running in a bronze bushing at moderate speed. Change it to H7/p6 and you've crossed into press-fit territory — the shaft is now larger than the hole, and assembly requires force or thermal expansion. Try to assemble the wrong shaft into that bore and you either seize the press or split the housing.

Three Categories of Fit — and Why They Exist

Clearance fits always leave a positive gap between shaft and hole. The smallest shaft within tolerance will still slide freely in the largest hole within tolerance. These are the fits for rotating bearings, sliding pistons, and anything requiring relative motion. The clearance isn't a mistake — it's a reservoir for lubricant, a compensation for thermal expansion, and the parameter that determines vibration behavior at speed.

Interference fits invert the relationship. The shaft is always larger than the hole across the entire tolerance range. Assembly requires pressing, heating, or chilling. The resulting joint derives its strength from the elastic deformation of both parts — the bore expands slightly, the shaft compresses slightly, and the resulting radial pressure creates friction that can transmit enormous torques without a single fastener. A locomotive drive wheel is pressed onto its axle this way. The interference is typically calculated in the 0.001 to 0.003mm per millimeter of shaft diameter range — meaning a 100mm shaft might have a 0.1 to 0.3mm interference. Get it wrong and the wheel either falls off or the axle cracks during assembly.

Transition fits occupy the probabilistic middle ground. Depending on where within their tolerance bands the actual shaft and hole dimensions fall, the assembled pair might have a small clearance or a small interference. These are used where location accuracy matters more than whether the parts can move, such as locating pins that need to be removable but positioned precisely.

The Geometry Problem: It's Never Just One Dimension

A shaft that's 25.000mm in diameter might still fail to assemble properly because tolerances apply to more than size. Geometric dimensioning and tolerancing (GD&T) extends the concept to form, orientation, location, and runout.

Cylindricity — the tolerance controlling how close a cylinder is to a mathematically perfect cylinder — matters enormously in bearing applications. A shaft could measure 25.005mm at both ends and 24.998mm in the middle. Nominally within tolerance on diameter, but the barrel shape will cause a rolling-element bearing to preload unevenly, generate heat, and fail prematurely. The cylindricity tolerance, often 0.005mm or less in precision spindle work, catches this.

Perpendicularity tolerances control whether a face is actually at 90° to an axis. A bolt flange that's out-of-perpendicular by 0.03mm across 50mm of diameter doesn't just leak gaskets — it induces bending stress into the bolt on every torque cycle, working toward fatigue failure in a way that no static calculation will predict.

DIY Reality: Why This Matters Outside the Factory

Home machinists and precision DIY builders encounter tolerance thinking constantly, even if they don't always name it that way. Building a linear motion system for a CNC router? The fit between a 20mm linear rail carriage and its rail is engineered to a specific preload — and if you buy mismatched components from different manufacturers using nominally the same specification, the carriages will either rattle or bind. The rail might measure exactly 20mm. The carriage bore might be exactly 20mm. But the dimensional chains don't sum the same way.

Working with reamed holes and precision dowel pins is another common encounter. A standard jobber drill nominally sized at 6mm typically produces a hole somewhere around 6.04mm to 6.08mm — drilling inherently produces oversized, slightly oval holes. A 6mm dowel pin held to an h6 tolerance (5.988mm to 6.000mm) will rattle in a drilled hole. To get a proper location fit, you drill undersize and ream to H7 — which for 6mm gives a hole of 6.000mm to 6.012mm. The pin then fits with controlled clearance and locates reliably.

This is why a 6mm reamer is not interchangeable with a 6mm drill for precision work, and why hardware-store pins are not interchangeable with ground dowel pins even when they share a nominal diameter.

Measuring for Real: Tools and Their Limits

Specifying a 0.02mm tolerance is meaningless if your measurement tools can't resolve it. Understanding measurement uncertainty is integral to tolerance work — a gauge with 0.01mm resolution and ±0.005mm accuracy can verify a 0.05mm tolerance band, but only barely. The general engineering rule is that your measurement system should have an uncertainty no greater than 10% of the tolerance being verified, which pushes you toward micrometers and air gauges rather than Vernier calipers for anything below about ±0.05mm.

Temperature matters at this scale. Steel expands approximately 11.7 microns per meter per degree Celsius. A 200mm shaft measured in a cold garage at 12°C will be about 0.022mm smaller than the same shaft measured in a 22°C machine shop — a difference that can move a borderline shaft from acceptable to reject without touching the part. The ISO standard for dimensional measurement is 20°C, and precision shops maintain temperature-controlled metrology rooms for exactly this reason.

Stack-Up: When Tolerances Multiply

Perhaps the most treacherous aspect of tolerance engineering is stack-up — the accumulation of individual part tolerances through an assembly. If five components each have a ±0.05mm tolerance on a dimension that chains through the assembly, the worst-case stack is ±0.25mm. For many applications that's fine. For a fuel injector with a 0.3mm nominal clearance, it means some assemblies will have negative clearance and won't function at all.

Managing stack-up drives designers toward tighter tolerances (expensive), fewer components in the chain (better design), or statistical tolerance analysis accepting that some small percentage of assemblies will be out of spec. Understanding this trade-off is genuinely engineering work — not just math, but judgment about what the application demands and what manufacturing can reliably deliver.

The Bottom Line

Tolerances are where the physics of the real world negotiates with the abstractions of engineering drawings. That 0.1mm — roughly four thousandths of an inch in imperial terms — isn't an arbitrary threshold. It's the scale at which surface films, thermal gradients, elastic deformation, and manufacturing variation all become players rather than footnotes. An engineer who internalizes this works with a fundamentally different mental model of how parts behave than one who treats all dimensions as exact numbers that get approximated in the shop.

The next time a part doesn't assemble, resist the urge to reach for the file. Reach for the micrometer first — then look at the drawing, and ask what fit was actually specified.