A 50 mm bore in an aluminum gearbox housing came back to us in a box with the bearing still inside it, and by the time we put a bore gauge in, the hole was 0.03 mm out of round. The note on the drawing said "press fit", in words, and the only dimension on the bore was 50 H7. Nobody had dimensioned the shaft, so the mating shaft was turned to 50 h6, which is 50.000 to 49.984 mm, because that is what a title block default produces when a feature is not called out. H7 against h6 is a sliding fit, not a press fit: the clearance runs from 0 to 0.041 mm. The outer ring of that bearing rotates under load, so it turned in the bore, fretted, and took the housing with it. Two thousand units were already in the field. *(Composite example; details changed.)*
We made both parts to print, which is the part of this story we still dislike. The parts were right, the assembly was wrong, and the reason was that the fit was never specified anywhere. The missing line cost a housing redesign, a rework instruction and a field repair, at roughly thirty times the value of the two parts involved. Everything below is about the four characters that prevent that phone call.
We are a CNC shop in Dongguan: 200+ machines on the floor, 23+ years of this work, ±0.01 mm on the features we quote, and IATF 16949 / ISO 9001 / ISO 13485 on the wall. Free DFM comes with the quote, and the quote comes back within 24 hours. None of that saves a drawing that says "press fit".
Hole Basis, Shaft Basis: Why H Is The Default
Two systems exist, and in the drawings that reach us, hole basis accounts for something like nine out of ten.
In the hole basis system the hole is always H, which means its lower deviation is zero and the hole is never smaller than nominal. You then move the shaft letter to get the fit you want: g6 for free running, h6 for sliding, k6 or n6 for location, p6 for a light press. In the shaft basis system it runs the other way. The shaft is h, with an upper deviation of zero, and the hole letter moves instead.
Hole basis wins on tooling, and that is the whole reason. A hole of a given size is made with a tool of that size, a drill, a reamer, or a boring bar set to a diameter, and those tools sit in the crib in standard sizes. A shaft can be turned to any diameter at all, because the operator moves the tool a few microns and takes one more pass. If you want four different fits on four different 50 mm holes, hole basis lets us do it from one row of reamers. Shaft basis would need four different reamer sizes at the same nominal diameter, and nobody keeps that in stock.
Shaft basis pays in two situations we see regularly:
- One long shaft carrying several different fits along its length. A gearbox input shaft with a bearing at one end, a gear in the middle and a coupling at the other is easy to turn in steps; boring four different hole sizes into four different parts is not. We cover this kind of part in our transmission shaft machining guide.
- Cold drawn shafting used as it comes. If the bar is already sized and you are not going to turn it, the shaft is the fixed member and it is cheaper to size the holes around it.
If you do not say which system you are using, we assume hole basis, and so does every other shop and every standard table you will look at.
The 50 mm Example, Worked Out
Shaft and hole fits are the shortest way to say what an assembly has to do. Every row in the table below is the same bore against a different shaft letter.
Here is the same 50 mm hole against five different shafts, all of it from ISO 286-1. The hole stays H7 in every row, 50.000 to 50.025 mm, a tolerance of 25 micrometers. That repetition is not laziness, it is the point of hole basis: one hole, one reamer, five fits by changing the shaft.
| Hole (H7 in every row) | Shaft | Shaft range at 50 mm | Resulting clearance or interference | Typical use |
|---|---|---|---|---|
| 50.000 to 50.025 | g6 | 49.991 to 49.975 | clearance 0.009 to 0.050 mm | free running, a collar that has to slide |
| 50.000 to 50.025 | h6 | 50.000 to 49.984 | clearance 0 to 0.041 mm | sliding and locating, parts taken apart by hand |
| 50.000 to 50.025 | k6 | 50.018 to 50.002 | 0.023 mm clearance to 0.018 mm interference | accurate location, assembled with a soft hammer |
| 50.000 to 50.025 | n6 | 50.033 to 50.017 | 0.008 mm clearance to 0.033 mm interference | location under load, assembled with a press |
| 50.000 to 50.025 | p6 | 50.042 to 50.026 | interference 0.001 to 0.042 mm | light press, gears, bearing seats with a rotating outer ring |
Read the p6 row twice, because it is the one that surprises buyers. The maximum interference is 0.042 mm, which sounds firm. The minimum is 0.001 mm, which is a thousandth of a millimeter on a 50 mm diameter, and on a rough mating surface that is barely more than the height of the finish peaks. A part at the loose end of p6 and a hole at the large end of H7 will go together with finger pressure and come apart in service.
The grades themselves, IT6, IT7 and where the numbers come from, are worked through in our ISO tolerance grades guide. The general tolerance rules for everything else on the part are in our CNC machining tolerance guide.
What To Write On The Drawing
This is the table we wish more prints carried. The middle column is the whole message of this article: write the fit symbol, not the English.
| Situation | Fit to write | What we do in the shop |
|---|---|---|
| Bearing outer ring rotating in the housing | 50 H7/p6, interference 0.001 to 0.042 mm | Ream or fine bore to H7, turn the seat p6, press on an arbor with the force recorded, and re-measure the bore on the first piece after pressing |
| Bearing outer ring stationary, inner ring rotating | 50 H7/h6, clearance 0 to 0.041 mm | H7 bore, h6 seat, hand assembly, so the bearing can be replaced in service without a press |
| Dowel pin location between two plates | 50 H7/n6 if the joint comes apart, 50 H7/p6 if it does not | Ream both holes in one setup wherever the design allows it, then press the pin with an arbor |
| Slip fit for adjustment, a collar or a knob that slides and locks | 50 H7/g6, clearance 0.009 to 0.050 mm | Turn the shaft g6, check that the parts slide under their own weight, and record both diameters on the FAI |
| Press fit for a gear on a shaft | 50 H7/p6, with the seat ground | Grind the seat, verify the interference on one instrument, and warm the gear or chill the shaft before pressing if the hub wall is thin |
Two remarks on that table. The bearing rows are not interchangeable, and getting them backwards is the most common fit error we see. A bearing with a rotating outer ring needs an interference fit, because a clearance fit lets the ring creep in the bore, and creeping produces fretting corrosion, then ovality, then noise, then the box on our bench. A bearing whose outer ring sits still while the inner ring rotates can run happily as a clearance fit, and a clearance fit makes service possible. We wrote about the parts around that decision in our automotive metal bearing guide.
Second, a fit callout never travels alone. The bearing seat also needs position and perpendicularity to the shaft axis, or the bearing goes in crooked and the careful fit is wasted. That side of the print is covered in our GD and T true position inspection guide, and it belongs in the same review as the fit.
"Press Fit" Is Not A Specification
When a drawing says "press fit" in words, we have to guess between 0.001 mm and 0.042 mm, and that is a range of forty to one on the thing that decides whether the part holds torque. We will not guess on a production part. We will ask.
What we need in order to answer is not a tolerance. It is: how much torque does the joint carry, what are the two materials, how thick is the wall around the hole, and what temperature does it run at. Give us those four and we will put a symbol and a number on the print during the DFM pass, at no charge, which is the cheapest conversation in this whole article. Our DFM analysis guide shows what else comes back in that review.
Thin Walls Move More Than The Fit You Asked For
An interference fit works by stretching the outer part. Put 0.03 mm of interference on a 50 mm bore in a chunky aluminum housing and the strain is 0.03 over 50, or 0.0006, and the resulting hoop stress is far below the yield of any aluminum we machine. The bore opens by about the interference you asked for, the joint grips, and nothing else moves.
Put the same 0.03 mm into a thin walled sleeve and the wall does not push back evenly. It deflects, and the bore can close by more than the interference you specified, so the shaft you pressed in is being gripped by a bore that has also moved inward somewhere else. On a thin section, an interference fit can distort the part more than the fit itself, and the first article is where that shows up.
This is why we ask for wall thickness before we quote an interference fit, and why we measure the bore after pressing on the first piece. If the geometry is marginal, the honest answer is often a different design: a thicker boss, a split hub with a clamp screw, or a key and a setscrew with a clearance fit instead of a press at all.
Temperature: The Fit You Lose At 40 °C
Aluminum grows at about 23 × 10⁻⁶ /°C and steel at about 12 × 10⁻⁶ /°C. The difference, 11 × 10⁻⁶ /°C, is what matters when an aluminum housing holds a steel shaft, and it moves in the direction most purchasers do not expect if they have only ever thought about steel on steel.
At 50 mm, a 40 °C rise opens the aluminum bore by 50 × 23 × 10⁻⁶ × 40, about 0.046 mm, while the steel shaft inside it grows by only about 0.024 mm. The pair gains roughly 0.022 mm of clearance. For a clearance fit that is harmless: an H7/g6 joint that starts at 0.009 mm minimum runs at about 0.031 mm when hot, so it gets looser, not tighter. For an interference fit it runs the other way, and it is the dangerous one. An H7/p6 pair loses the same 0.022 mm from a band of only 0.001 to 0.042 mm, and at the loose end of that band the joint is no longer an interference fit at all. Size the fit at the temperature it will run at, not at the inspection bench.
Housings run warm. Amplifier chassis, gearboxes, anything with a motor bolted to it, all of them sit above room temperature once they have been running for an hour. Tell us the operating temperature and we will size the fit at temperature rather than at the inspection bench.
Coatings Eat Clearance
Hard anodizing grows roughly half of its thickness outward and half inward. A 50 micrometer coating therefore adds about 25 micrometers per surface, and 25 micrometers is 0.025 mm, which is nearly three times the 0.009 mm minimum clearance of an H7/g6 sliding pair. Machine a shaft to g6, coat it, and the free running fit becomes interference.
There are three honest fixes, and they have to be chosen before the parts are made: mask the fit surface so it stays bare, machine the shaft undersize by the expected growth, or open the hole to suit. What does not work is discovering it at assembly. The full behavior of the coating, including the thickness we can actually hold in production, is in our hard anodizing guide.
How We Check A Fit Nobody Can See
You cannot verify a press fit with calipers, and we are going to keep saying this until it stops coming up in emails. The whole interference at 50 mm H7/p6 runs from 0.001 to 0.042 mm. A caliper in a human hand is doing well to resolve 0.02 mm, so the instrument is roughly twenty times coarser than the feature, and two readings taken with it and subtracted from each other are not a measurement.
The rule we work to: both diameters are measured on the same instrument, by the same operator, in the same session. A micrometer set, an air gauge, or the CMM, one of them for both members. That way the instrument error largely cancels and the difference between the two readings means something. When the numbers are disputed, the gauge has to clear a gage R and R study first, because arguing about 0.005 mm with a gauge that contributes 0.008 mm of its own variation is a waste of everybody's afternoon.
On production parts this runs through the inspection plan we already use for everything else: bore and shaft both recorded on the first article inspection, method documented in our CNC inspection methods and CMM guide. Shaft seats that need it are ground, and the turning side of that work is described in our CNC turning service guide.
What buyers actually ask us
"Do I really need H7, or will ±0.05 be cheaper?"
Usually the tolerance itself is not what costs money, the process step is. A 50 mm hole at ±0.05 comes off the drilling and boring cycle you are paying for anyway; H7 needs one more operation, a ream or a finish bore, and about two minutes of cycle time. The expensive consequences show up later, in a bearing that will not go in, or in a fit you now have to inspect on every piece to prove it still works. We would rather quote H7 and remove the doubt.
"The shaft is cold drawn bar. Can I use it as it comes?"
Yes, and it changes which system you should be writing. If the bar is already at its finished size and you are not going to turn it, you are running shaft basis whether you call it that or not, and the hole has to be sized to the bar you actually bought, not to the nominal. Tell us the bar spec on the print and we will work from it.
"It is an aluminum housing on a steel shaft and it runs hot. Does that matter?"
It matters by about 0.022 mm of gained clearance for every 40 °C of rise at 50 mm, more than twice the minimum clearance of a sliding fit and more than half the maximum interference of an H7/p6 pair. Give us the operating temperature and the assembly temperature, and we will put the fit at the condition it actually runs in. This is the single most common omission on drawings that come to us from outside the automotive world.
"My drawing is from a US customer and says RC6. Can you work to that?"
Yes. ANSI B4.1 and the ASME classes, RC, LC, LN and FN, are a different way of writing the same idea, and we can map them across to ISO. What we ask is that you tell us which system the drawing uses before we quote, because a drawing that mixes an ASME class on one feature with H7 on another usually means two people edited it in different years. We dimension to ASME Y14.5 conventions, which you can read about at ASME. For context on where this kind of print shows up, see our automotive industry page.
"Why is the interference bore more expensive than the plain bore next to it?"
Because of what has to happen after the chip is cut. The seat gets ground rather than turned, both members get measured on one instrument before assembly, the press force gets recorded, and the first piece gets the bore measured again after pressing to prove the wall did not move. None of that is exotic, but it is five extra steps, and it is the difference between a fit that holds and a fit that looks right on the inspection report.
FAQ: shaft and hole fits
What is the difference between a clearance fit and an interference fit?
A clearance fit always leaves a gap between shaft and hole, so the parts assemble by hand and move relative to each other in service. An interference fit always leaves the shaft larger than the hole, so the parts have to be pressed, heated or chilled to go together, and the resulting contact pressure is what carries the load. A transition fit can give either, depending on where the two parts fall in their tolerance bands.
What does H7 mean on a drawing?
H tells you where the band sits: for a hole it means the lower deviation is zero, so the hole is never smaller than nominal. The 7 tells you how wide the band is, and that width is set by size step, not by opinion. Write both characters and a shop anywhere knows both facts without opening a calculator.
How do I say press fit without writing the words press fit?
Write H7/p6 and the interference range arrives with the symbol, somewhere between finger pressure and an arbor press. Write the words and somebody downstream decides which one you meant, usually at 4 pm on a Friday.
How do you prove the fit was right after the parts are pressed together?
You do not measure the joint, because it is not coming apart to be measured. We measure the hole and the shaft separately, on the same instrument, before assembly, record both numbers on the FAI, and the press force and the post press bore check on the first piece confirm that the wall behaved. After that the fit is guaranteed by the two recorded diameters.
If you have a shaft and a hole that have to carry torque, hold position, or come apart again next year, send us the drawing and tell us what the assembly actually has to do. We will come back in plain language, tell you whether that bore wants H7 or whether you can open it up and skip the ream, and we will put a number on the print instead of the word "press fit". Quote within 24 hours, DFM free, and if you need the shaft side as well, our CNC turning service is where that work runs. Reach us on WhatsApp at +86 17620300785 or by email to gongtianbao@xhlmarketing.com, and we will answer in plain language. Talk to our team