Why I stopped trusting “just make it H7”
I’ve been specifying fits on rotating assemblies for close to twenty years, and the thing that still surprises me is how fast smart engineers reach for H7 without asking why. It’s not their fault. Every textbook dimensioning chapter pushes base-hole practice, every CAD template auto-fills H7 next to a diameter symbol, and most tolerance charts only bother to show the H7/g6 and H7/p6 rows. So the 32 mm shaft gets g6 because “that’s what the fit table says,” and nobody stops to ask whether someone will ever turn this part with a wrench, whether it carries an oil film, or whether it spins at 6,000 rpm under a 2.3 kW motor.
Let me give you a concrete failure. Six years back I signed off on a drive sprocket meant to be a medium-interference press fit on a 28 mm motor shaft. The drawing said H7/p6, defensible on paper — roughly 1 to 35 µm of theoretical interference depending on where each part lands in its tolerance zone. We pushed it on with a hydraulic press, torqued it down, and it ran fine on the test stand. Then production batches came back: plain cylindrical bores, no lead-in chamfer, and a supplier who ground every shaft to the high end of p6 and reamed every bore to the low end of H7. The result wasn’t 35 µm of interference, it was a sprocket that needed 9 tonnes of force to move and cracked its keyway hub the third time it was pulled for maintenance. Nothing on the drawing was wrong. The tolerance band selection was textbook. The environment around it was not.
The drawing is not the fit. The drawing is a promise about how two parts will cooperate once machining, temperature, surface roughness and maintenance all have their say.
Reading ISO 286 without the headache
Tolerance bands in ISO 286 are a two-letter-and-number code: the capital letter is the fundamental deviation of the hole, the lowercase letter is the fundamental deviation of the shaft, and the number is the tolerance grade. The letter tells you where the band sits relative to the zero line; the grade tells you how wide the band is. That’s the whole trick. H means the hole’s lower deviation is zero, so an H7 hole runs from 0 to +21 µm at 30 mm nominal. g6 sits below the zero line, from about −7 to −20 µm at the same size. Subtract the two and you get a clearance window of 7 to 41 µm — a sliding fit that will actually slide.
| Fit type | Typical code | Ø30 interference / clearance | Where I use it |
|---|---|---|---|
| Clearance / sliding | H7/g6 | +7 to +41 µm | Spindles, guide pins, parts that must be hand-assembled and free to slide |
| Clearance / running | H7/f7 | +20 to +54 µm | Bearing journals with an oil film |
| Transition / snug | H7/k6 | −2 to +23 µm | Gears and pulleys that are keyed but need light hand pressure |
| Interference / light press | H7/p6 | +1 to +35 µm | Bushings, sprockets, parts that must not move under load |
| Interference / heavy press | H7/s6 | +21 to +55 µm | Steel hubs on solid shafts, shrink fits |
Two habits will save you pain. First, stay in the base-hole system unless you have a real reason — ground bar stock and finished shafts are cheaper to alter than bores, so put the bigger machining tolerance on the hole. Second, write the actual deviation numbers on the drawing instead of trusting the reader to recall them. I’ve walked onto assembly lines where the fitter had a tape measure and a chart photocopied so many times the H7 row was unreadable.
A four-step way to pick a fit (one I actually use)
- Write down the job the joint has to do. Is it locating, transmitting torque, or carrying a load while allowing movement? If torque has to cross the joint, you are either planning a key or spline, or you are planning interference — and you should write the torque number down before touching a tolerance table.
- Decide the disassembly question now, not later. Will this ever come apart on a customer site without a press rig? That single question eliminates half the argument. A field-replaceable bushing means sliding or light press. A permanent hub means heavy press or shrink.
- Estimate from the table, then sanity-check the extremes. Take the worst-case clearance and the worst-case interference, not the nominal. If the worst-case clearance means a wobble you can feel, the fit is wrong even when the nominal looks fine.
- Write the numbers, add the chamfers, and tell the machinist what the intent is. A 0.5 mm × 30° lead-in chamfer on the bore and the shaft end turns a brutal press into a routine one.
Here is where I differ from a lot of design guides. Most of them stop at the table. I insist the table is only the start, because press fits are where theory and shop reality go to war. The theoretical interference on paper gets eaten by surface roughness peaks, by the elastic flattening of the chamfer, and by temperature differences if you shrink a steel hub onto an aluminium shaft — an 80 °C temperature swing moves a 40 mm diameter by about 0.04 mm, which is more than the entire interference allowance of some of these fits.
Press force you can actually quote
The force to press a plain cylindrical joint together, ignoring friction variation, is roughly F = π · d · L · p · μ, where d is the nominal diameter, L the engagement length, p the radial contact pressure from the interference, and μ the coefficient of friction. For clean steel-on-steel with a light lubrication wipe, μ lands around 0.1 to 0.15. Let me use realistic numbers so the formula means something. Take a 30 mm hub pressing onto a shaft with 25 mm of engagement and 20 µm of effective interference. For that interference on a steel hub you get a contact pressure of roughly 60 MPa. That puts the press force at about π × 0.03 × 0.025 × 60 × 10^6 × 0.12 ≈ 17 kN. A domestic hydraulic press handles that comfortably. Double the interference and you are suddenly quoting 35-plus kN and wondering why the operator is leaning on the lever with both hands.
Surface finish matters far more than most textbooks admit. Two parts at the same nominal interference, one ground to 0.4 µm Ra and the other turned to 3.2 µm Ra, behave completely differently — the roughness peaks interlock and add hidden interference, and wringing, galling and seizure all become plausible during assembly. If you are doing heavy press fits and parts keep sticking halfway through the push, check the surface finish before you blame the tolerance band.
One more honest warning about temperature. Shrink fits are wonderful right up to the moment the workshop does not have the oven, the induction heater or the dry ice. I once specified a shrink fit as “the clean way” on a project and then watched the machinist argue with my drawing because he had to heat the part over a gas ring. The design was right; the shop-floor logistics were not. If your customer’s maintenance crew does field rebuilds, a press fit or even a sliding fit with a locking adhesive will serve them better than a doctrine-perfect shrink fit.
One field story worth reading twice
Last year a customer’s indexing table locking pin kept seizing. Twelve millimetre diameter, H7/k6, steel pin in a cast housing — a transition fit that should have been snug but free. The drawing was right, the finish was right, and yet the operator needed a soft-face hammer and swearing to move it. We found it after three phone calls: the housing bore was being reamed with a worn fixture that walked off centre by about 8 µm, pushing the pin’s actual clearance to nearly zero on one side while the pin still measured perfectly to print. The tolerance band was never the problem. The process capability was. That is the sentence I want every design engineer to tattoo somewhere visible: a fit is only as good as the process that makes it.
It is also why I have stopped obsessing over chasing the narrowest IT grade available. If your shop routinely holds IT6 on that spindle with reaming, great — use it. If the same shop drifts by 10 µm from shift to shift, then buying IT6 on the drawing achieves nothing except more rejected parts. Pick the grade the process can actually hold, then choose the deviation so the working window lands in the right place. That ordering never appears in the textbooks but it decides whether your fit survives contact with a third shift.
The mistakes I keep seeing (and the fixes)
| Mistake | What actually happens | What I do instead |
|---|---|---|
| Copying the chart default H7/g6 for every job | Sliding fits that rattle under vibration, or ends up too tight for hand assembly | Confirm the job — locate, transmit, or slide — before choosing anything |
| Specifying interference without a chamfer | Scored bores, galling, and a press operator who hates your drawing | Add a 0.5 to 1 mm lead-in chamfer on bore and shaft end |
| Ignoring the worst-case corners | Nominal looks perfect but the maximum-material pair seizes or wobbles | Always check the max-metal and min-metal combinations by hand |
| Reusing a fit from a different nominal size | A fit that was lovely at 80 mm is a different animal at 12 mm | Re-check the deviation values every time the size changes |
| Leaving out finish and assembly method notes | The assembly floor guesses, gets it wrong, and blames tolerancing | Put finish, chamfer and press or shrink method on the drawing |
A short glossary for the next conference call
| Term | What it means |
|---|---|
| Base hole system | The hole keeps H (zero lower deviation); the shaft does the adjusting |
| Base shaft system | The shaft keeps h; the hole adjusts — used with ground bar stock |
| Fundamental deviation | The letter: where the band sits relative to the zero line |
| IT grade | The number: how wide the band is, IT5 tight to IT11 loose |
| Maximum material condition | The extreme where each part takes on the most metal |
| Wringing fit | Interference so tight the parts bond at assembly |
How I record fits so nobody has to rediscover them
Early in my career I treated every fit as a fresh decision and it cost me, because three designs later I would pick a different combination for exactly the same joint and create two parts that did not match the stock on the shelf. Now I keep a one-page internal standard per product family. Thirty millimetre shaft ends get H7/k6 with a 1 mm chamfer and 1.6 µm Ra on the journal, full stop. 100 mm hubs get H7/p6 with a note that they are only assembled with a press. Nothing glamorous — but after two years the machinists quote from memory and the rework count for fits quietly drops to near zero. You do not need fancy tolerance software to get this. You need consistency and someone who is willing to make a decision and write it down.
Closing thoughts from a guy who has pushed too many hubs on and off
Fits are not decoration on the drawing. They are the contract between design intent and what a fitter can actually achieve in a dirty workshop on a Tuesday afternoon. Ask the four questions — function, assembly, disassembly, environment — and the ISO letters stop feeling like a magic alphabet and start reading like plain language. And when you are in doubt, write the deviation numbers out longhand and walk the print to the machinist. It takes sixty seconds and has saved me more rework than any tolerance software ever did.
Honestly, the best fit is the one the shop can hit reliably, not the one that scores the tightest on a simulation screenshot. That trade-off never makes it into the marketing brochures, but it is the whole game. If this article saves you one seized hub or one sleepless night before a customer visit, it did its job.
