A constant-section bearing can spin freely on the bench and then tighten up once the housing is bolted down. The bearing may still be within specification. A seat that is slightly out of round, an interference fit that is too tight, or uneven clamping can each change how a thin ring behaves after mounting.
There is no single bearing fit tolerance that suits every constant-section bearing. Choose the fit from the model's dimensional limits, its internal clearance or preload, which ring the load rotates against, and the operating temperatures. Then check the seat geometry and the mounted behavior.
Naming a fit such as H7 or k5 doesn't finish the job. On a 50.8 mm (2 in) shaft, k5 puts roughly 2–30 µm of interference on KA020CP0, whose published unmounted radial clearance is only 0–12.7 µm. The two numbers measure different things, but the gap between them is the reason to look closely.
This guide covers the shaft and housing interfaces shown in Figure 1. For bearing types and general selection, see our thin-section bearing selection guide.
Key points
- A fit tolerance is a range, not a number. On KA020CP0's published limits, k5 interferes by about 2–30 µm, while h5 runs from loose to tight.
- Which ring needs retention depends on how the load moves relative to it. A ring under stationary load may take a clearance fit.
- A seat can pass a diameter check and still distort a thin ring. Measure form as well as size.
- Check the assembly at four points: before mounting, after seating, after final clamping and under load.
Why constant-section bearings need a closer look
Within the broader thin section bearings category, constant-section series keep the same cross-section as the bore size increases. The KA series (1/4 in × 1/4 in section) keeps the same 0.25 in (6.35 mm) radial section whether the bore is 2 in or 4 in. That is 12.5% of the bore diameter at 2 in and 6.25% at 4 in.
Rings this thin tend to follow the shape of whatever supports them. An out-of-round seat can push into the raceway, and an uneven shoulder can tilt the ring.
Three separate checks cover this:
Size: do the shaft and housing diameters give the intended clearance, transition or interference?
Form: are the seats round, straight and aligned with each other?
Mounted behavior: does the finished assembly meet its clearance, torque and runout limits?
A bore that passes the diameter check has answered only the first.
Start with which ring the load rotates against
What matters is whether the load direction moves around the ring. Take a rotating shaft with a radial load fixed to the machine frame. The inner ring sees a rotating load and the outer ring sees a stationary one. If the housing rotates instead, with a fixed shaft and the same fixed load, the roles swap.
A ring under rotating load needs enough retention to stop creep, where the ring slowly turns on its seat and wears it. The usual remedy is interference, which in practice means a press fit. A ring under stationary load can often take a clearance or light fit, depending on how it must be located. Reversing loads, shock and uncertain load direction need their own review.
On a thin ring, more interference can also change the raceway shape and reduce internal clearance. Don't keep tightening the fit until creep stops without checking clearance or preload. Confirm the bearing configuration first: a radial bearing, a four-point contact bearing and a matched angular-contact pair can need different mounting arrangements.
Reading H7 and k5 as numbers
H7 and k5 are ISO 286 tolerance zones. H7 is a hole (housing) zone and k5 is a shaft zone. The examples below use KA020CP0, whose nominal bore is 2 in (50.8 mm) and nominal outside diameter is 2.5 in (63.5 mm). Both fall in the 50–80 mm size step of the standard (above 50 mm, up to and including 80 mm). The limit deviations are:
ISO 286-2:2010 limit deviations for the 50–80 mm size step. They are dimensional limits, not a mounting recommendation for this bearing.
The bearing ring has its own published limits, and they stack on top of these. That stack is the real fit range.
One sign rule for both seats
Use the same rule at both interfaces: diametral interference = diameter of the inner part − diameter of the outer part. Positive means interference. Negative means clearance.
At the shaft, the inner part is the shaft and the outer part is the bearing bore.
At the housing, the inner part is the bearing outside diameter and the outer part is the housing bore.
Imax = inner part maximum − outer part minimum
Shaft and bearing fit tolerance: the inner-ring example
KA020CP0 is published with a bore tolerance of −0.0006 in to 0, which converts to −15.24 µm to 0, so the bore runs from 50.78476 to 50.80000 mm. Its published radial clearance is 0–0.0005 in (0–12.7 µm).
The shaft limits come from the ISO zones above. This is an arithmetic example, not a LILY fit recommendation.
Take k5. I_min = 50.80200 − 50.80000 = +0.00200 mm, and I_max = 50.81500 − 50.78476 = +0.03024 mm. The 28.24 µm spread is the 13 µm shaft zone plus the 15.24 µm bore tolerance.
The maximum interference, 30.24 µm, is more than twice the largest published unmounted radial clearance of 12.7 µm. The two numbers aren't interchangeable, because only part of the interference reaches the raceway, and how much depends on ring dimensions, shaft stiffness and surface condition. Still, the comparison shows why k5 can't be copied from a general table.
h5 has the opposite problem. Its range includes loose assemblies, and a loose ring under rotating load can creep. Which zone is right, if either, is a question for the model drawing and an engineering review.
One assembly that feels right doesn't validate the drawing range. For production design, check the extreme combinations. If parts are selectively matched, say so on the drawing instead of assuming every shaft fits every bearing.
Housing-to-outer-ring example
KA020CP0 is published with an outside diameter tolerance of −0.0005 in to 0 (−12.7 µm to 0), so the bearing OD runs from 63.4873 to 63.5000 mm. With an H7 housing (63.500–63.530 mm):
I_min = 63.4873 − 63.530 = −0.0427 mm (42.7 µm clearance)
I_max = 63.500 − 63.500 = 0.000 mm (line-to-line)
So the outer ring sits between line-to-line and 42.7 µm loose. Figure 2 plots both calculated ranges. For a stationary load that can be acceptable, provided the housing bore stays round and the ring is located and retained as the design requires. Print the sign rule on the inspection sheet so that "+10 µm" means the same thing to the designer and the inspector.
Is there a bearing fit tolerance chart for constant-section bearings?
Not a universal one, and that includes any bearing press fit tolerance chart. Catalog charts assume a particular bearing type, ring thickness, support material and load condition. Some assume steel shafts and cast-iron or steel housings, and a thin ring in an aluminum housing breaks that assumption.
Use a chart for a first estimate. Then confirm against the drawing for your exact part number: the dimensional limits, the internal clearance or preload, and the permitted seat geometry. The section called "Get the fit reviewed before you release drawings" lists what to send us for that check.
Check seat geometry before changing the fit
Measure form as well as diameter. As a first shop-floor screen, take readings at three axial positions and in two perpendicular directions at each, six in all. Those readings can reveal taper and obvious ovality. They don't certify roundness or cylindricity. Figure 3 shows what a distorted seat can do to a thin ring.
Two-point diameter readings can miss odd-lobed shapes, including three-lobed ones, so similar readings don't prove a seat is round. For a specified roundness tolerance, use a method that can evaluate it. NIST describes roundness measurement from full 360° traces, including the effect of instrument spindle error.
Also check shoulder flatness, shoulder runout and the alignment of separate seats to the drawing datums. Shoulders and clamp faces should contact only the ring they support, without fouling the other ring, the cage or a seal. Published ring-face diameters are useful references, but they aren't approved shoulder or clamp limits. Take those from the assembly drawing. Make sure the corner fillet clears the bearing chamfer: a ring resting on the fillet can look seated while its face is unsupported.
If you can run only one check first, compare the housing bore before and after tightening the fasteners. For a split housing or flexible mounting plate, do it in the assembled state the drawing defines. If tightening pushes the bore outside its limits, a new bearing won't fix the support problem.
If rotation changes after clamping, compare the support geometry. Record the housing bore and a rotation check before and after tightening, under the same conditions. A change points inspection toward clamp contact, fastener sequence or housing stiffness. It doesn't establish the cause. Don't convert a measured out-of-roundness value directly into an internal-clearance reduction.
Temperature and axial growth
Dimensions only compare on a common reference. The standard reference temperature is 20 °C, as NIST explains in its paper on the 2016 revision of ISO 1. When you check a close fit, record part and gauge temperatures.
For a simple free-expansion estimate, use ΔD = α × D × ΔT, where α is the linear expansion coefficient, D the diameter and ΔT the temperature change.
These values are illustrative, not certified properties of any LILY model. Take an aluminum housing (α ≈ 23 × 10⁻⁶ /°C) around a steel outer ring (α ≈ 12 × 10⁻⁶ /°C), with a 63.5 mm diameter and a 40 °C rise.
(23 − 12) × 10⁻⁶ × 63.5 mm × 40 = 0.02794 mm, or 27.94 µm
That is close to the 30 µm width of the H7 zone. With these assumptions (equal temperatures, free expansion), heating opens the housing away from the ring and cooling closes it back. Whether cooling produces interference depends on the starting fit.
KA020CP0's page lists an operating range of −30 °C to 110 °C. Apply the same coefficients to that range, purely as an illustration. Cooling from 20 °C to −30 °C reduces the diameter difference by 34.925 µm. Heating to 110 °C increases it by 62.865 µm.
A housing that starts at the 42.7 µm clearance limit still has about 7.8 µm left at −30 °C. One that starts line-to-line ends up with about 34.9 µm of interference. These are fit-change estimates, not measured interference or approved assembly temperatures. Coefficients vary with alloy and temperature, and constrained supports or unequal component temperatures need their own assessment.
Along the shaft, the same steel coefficient gives 12 × 10⁻⁶ × 500 mm × 40 °C = 0.24 mm (240 µm) for a 500 mm shaft. The housing and supporting structure grow too, so compare the growth of each. Decide which support locates the shaft axially and where relative movement is allowed. If that movement is restrained, unintended axial load can develop, and its size depends on the stiffness of the assembly.
A clearance fit alone doesn't guarantee that a ring can float, because shoulders, clamp faces, seals and seat friction can hold it. Don't loosen a clamp that sets preload just to create movement.
Verify the assembly in four stages
Repeatable conditions let each measurement answer one question. The last column shows where to look first if tightness first appears at that stage. These are inspection priorities, not diagnoses.
Set acceptance limits before you test, and record the fixture, load, speed, temperature, lubrication and instrument uncertainty. The same torque reading can be fine in one mechanism and excessive in another. Starting torque and running torque are different checks. A repeatable tight spot points toward geometry, but contamination and bearing damage can feel the same, so don't diagnose from a hand-spin test alone.
When you press a ring on, apply the force to the ring being fitted, through tooling that supports its face evenly. Never push through the balls. For thermal mounting, use the product-approved temperature limits. The operating temperature rating isn't automatically a safe mounting temperature.
Build the data sheet from the model page
The CS category page lets you screen models by dimensions, load direction, materials and sealing. The model page then gives the fields a fit review starts from. Here is KA020CP0, with inch values converted using the exact factor of 25.4 mm per inch (NIST SP 811, Appendix B.8):
LILY's variant information says clearance and preload values apply to unmounted bearings under zero measuring load. The mounted value can differ with fit, clamping and temperature. Also confirm whether the drawing defines size as a mean diameter, a local size or something else before you use catalog limits for inspection.
The last digits of the part number matter just as much. In an inch-series number, position 6 is the bearing type, position 7 the cage, position 8 the tolerance class and position 9 the clearance or preload. KA020CP0 reads as a radial-contact bearing (C), with a one-piece snap-type cage (P, brass on this model), tolerance class 1 / ABEC 1F (0) and standard clearance.
It has no ninth-position letter, so use its published clearance and don't infer one from the tolerance digit. A different eighth digit can change dimensional tolerance, run-out or both, and a ninth letter selects a specific clearance or preload range. Decode the full designation against the variant table.
Bore alone doesn't identify the mounting envelope either. Two radial models on the category page share a 1 in bore and a 1.375 in outside diameter, yet differ in width, sealing and ring-face diameter:
These are nominal dimensions, not machining limits, and the ring-face conversions are rounded to 0.01 mm. The different seal and ring-face arrangements matter when you check tooling and clamp clearance, even though bore and outside diameter match. Add the bearing tolerances, internal clearance or preload and seat requirements from the model's current drawing before you release shaft and housing drawings. To see every model, browse the Constant Section (CS) bearing range.
Get the fit reviewed before you release drawings
For a model-specific fit review, contact us with:
the exact part number, including any clearance or preload suffix
shaft and housing limits and materials
which ring the load rotates against, and the load directions
the operating temperature range
your mounted acceptance criteria for clearance, torque and runout
measured seat geometry and the tightening and measurement records, if you are investigating an existing assembly
Current bearing and installation drawings help if you have them. The review should settle the bearing configuration, the allowed interface ranges, the residual clearance or preload, and how you will verify them.
FAQ
Can I use retaining compound to make up for a loose seat?
Only as part of an approved repair or assembly design. A compound adds its own gap, curing and temperature requirements. It doesn't restore an out-of-round bore or correct misalignment.
Should I measure the housing before or after anodizing or plating?
Measure the final functional surface, and make the drawing state whether the tolerance applies before or after the finish. Don't assume a fixed bore change per micron of coating. Buildup, substrate conversion, masking and process variation all matter.
Can I reuse a thin-section bearing after pressing it off an interference seat?
It depends on the part, the fit and how it was removed. Send our engineers the part number and the mounting history for a case-specific answer. What to include is listed in "Get the fit reviewed before you release drawings" above.
Are seat clearance and internal clearance the same thing?
No. Seat clearance is the gap between a ring and its shaft or housing. Internal clearance is the free movement between the inner and outer rings with no load applied, and LILY publishes it for the unmounted bearing. A loose seat tells you nothing about internal clearance after clamping, and an interference seat can remove some of it.
About the Author
