O-Ring Groove Machining: Tolerances That Decide Whether the Joint Seals
An O-ring is a nearly incompressible volume of rubber. It seals because the groove squeezes it by 15 to 30 percent, forcing the material to fill the gap between the two mating surfaces. Every groove dimension exists to control that squeeze, and every sealing failure traces back to one of four things: the groove is too deep, too shallow, too narrow, or too rough in the wrong direction.
That is why ±0.05 mm on a groove is not fussiness. It is the difference between a joint that holds 10 bar for a decade and one that weeps at 0.5 bar on the bench.
Squeeze, fill and the numbers behind the groove
Squeeze is the percentage of cord diameter lost when the joint closes: (cord − gap)/cord. For a static face seal, 20 to 30 percent is normal. For a radial static seal, 15 to 25 percent. For a dynamic seal — a piston or rod that moves — 10 to 20 percent, because friction and wear punish over-squeeze.
Depth sets squeeze and is the dimension to tolerance tightly: ±0.05 mm for cord up to 3.5 mm. Width is set by fill ratio — ring area must stay under about 85 percent of groove area, leaving room for thermal expansion and for swell in oil. A 2 mm cord in a 2.4 mm wide, 1.5 mm deep groove runs about 78 percent fill.
Too narrow and the ring pinches at the parting line; too wide and it moves until it extrudes into the clearance gap under pressure.
Face seal versus radial groove
A face groove is cut into a flat surface and seals axially against a cover. It is easy to machine and easy to inspect with a depth micrometer, but the mating face must be flat: 0.02 mm of flatness error on a 60 mm flange loses contact on one side. Bolt preload distorts thin covers, so a face groove in a 5 mm plate is a design risk before it is a machining one.
A radial groove is cut into a bore or a shaft and seals against a cylinder. It tolerates axial stack-up far better, which is why hydraulic ports and sensor bodies use one. Its weakness is concentricity: 0.1 mm of eccentricity on a 2 mm cord swings squeeze between 20 and 30 percent, the top of the band on one side and the bottom on the other.
Call out concentricity of the groove to the bore at 0.05 mm and it becomes a non-issue.
See also: Battery Housing Machining: Gasket Grooves, Flatness and Leak Test Prep and Sensor Housing Machining: Wall Thickness, Sealing Faces and Thread Coaxiality.
Surface finish and the direction of tool marks
Floor finish depends on seal type, and specifying the finest available finish is wrong about half the time.
Static seals work at Ra 1.6 to 3.2 µm: a rougher floor gives the elastomer something to grip, while a polished floor lets the ring creep and slide out of place during assembly. Dynamic seals want Ra 0.4 to 0.8 µm with Rz under 2.5 µm, because peaks abrade the ring on every stroke.
Direction matters more than roughness value. A continuous helical tool mark that crosses the groove from the pressure side to the atmosphere side is a leak path no elastomer can fill. Finish the groove floor with a radial plunge or one face cut, not a turning pass travelling across the width. Circular marks that follow the groove are harmless; a spiral that crosses it is not.
Burrs are the other silent killer: a burr on the low-pressure lip acts as a knife and shaves the ring at assembly, and the leak shows up a week later at the customer. Deburr by hand with a 0.2 mm radius scraper, never a wire brush dragged across the groove.
Groove dimensions by cord diameter
The table gives the nominal geometry we use for static seals at 25 percent squeeze. Depth tolerance is the callout that matters; width tolerance is normally +0.1/−0 mm.
| Cord diameter | Groove width | Groove depth | Depth tolerance | Min corner radius | Floor finish |
|---|---|---|---|---|---|
| 1.5 mm | 1.9 to 2.0 mm | 1.15 mm | ±0.05 mm | 0.2 mm | Ra 1.6 µm |
| 2.0 mm | 2.4 to 2.5 mm | 1.50 mm | ±0.05 mm | 0.2 mm | Ra 1.6 µm |
| 2.5 mm | 3.0 to 3.2 mm | 1.90 mm | ±0.05 mm | 0.3 mm | Ra 1.6 µm |
| 3.0 mm | 3.6 to 3.8 mm | 2.25 mm | ±0.05 mm | 0.4 mm | Ra 1.6 µm |
| 3.5 mm | 4.2 to 4.4 mm | 2.60 mm | ±0.08 mm | 0.4 mm | Ra 1.6 µm |
| 5.0 mm | 6.0 to 6.3 mm | 3.75 mm | ±0.08 mm | 0.6 mm | Ra 1.6 µm |
For dynamic service, add 0.1 mm to depth on cord of 3 mm and above and tighten the floor to Ra 0.4 to 0.8 µm. For vacuum service, reduce depth by 0.05 mm so squeeze rises toward 30 percent, and keep fill ratio under 75 percent.
The rejection list
Five causes account for nearly every failure.
- Shallow groove. Squeeze climbs past 35 percent, assembly force rises, the ring takes a compression set in storage, and a leak appears at low temperature when the elastomer stiffens.
- Deep groove. Squeeze falls below 10 percent, contact stress drops under the fluid pressure, and the joint weeps at exactly the pressure the datasheet promised.
- Helical tool mark. Invisible to the eye, visible on a leak test, and always at the worst moment.
- Burr on the low-pressure lip. Ring is cut at assembly; leak appears after a few thermal cycles.
- Groove eccentric to the bore. Squeeze varies around the circumference; the joint holds on the bench and fails after the first pressure spike.
The fix is always on the drawing: depth with a tolerance, groove-to-bore concentricity, corner radius, floor finish with a direction note, and a deburr requirement. Five lines of text that eliminate most arguments after delivery.
Machining sequence notes
Cut grooves after the bore and before coating. Anodize grows into a groove by roughly the coating thickness per side; on a 1.5 mm deep groove, 25 µm of Type III costs 3 percent of depth and moves squeeze by 2 points. Mask the groove, or cut after coating with a diamond insert.
On parts with several grooves, cut them in one setup from one datum. Three radial grooves turned in a single operation on a CNC turning center hold spacing to 0.02 mm; two operations double that and add a re-chuck mark on a sealing surface.
In a milled pocket, interpolate the groove with a tool smaller than its width and enter by helix so the entry mark lands on the pressure side; a plunge entry leaves a witness mark that behaves like a nick.
Send the O-ring part number with the drawing. Cord tolerance, hardness and swell in the actual media shift depth by a few hundredths, and a shop that knows the ring can compensate before cutting metal. Send the file here for a quote.
Frequently asked questions
How deep should an O-ring groove be?
Seventy-five percent of cord diameter for a static seal at 25 percent squeeze, so 1.5 mm for a 2 mm cord. Hold ±0.05 mm. Go 0.05 mm shallower for vacuum and 0.1 mm deeper for dynamic service on larger cord.
Why does my joint leak at low pressure but hold at high pressure?
Low-pressure leaks usually mean too little squeeze: the ring never develops contact stress above fluid pressure. Check depth first, then finish. A groove 0.1 mm too deep on a 2 mm cord loses about 5 points of squeeze.
Should the groove floor be polished?
No. Ra 1.6 µm holds the ring in place for static service. Polishing below Ra 0.4 µm encourages creep in face seals and removes the micro-texture that retains lubricant in dynamic seals.
Can the groove be cut after anodizing?
Yes, and it is often the better route: mask the groove during anodizing, or cut it afterwards with a diamond insert and accept a shorter tool life. Cutting before coating costs you 25 µm of depth per side and a squeeze change you cannot inspect.
Related reading
Send a drawing for a quote and DFM review.
STEP, IGES, X_T, PDF or DWG up to 50 MB. Tell us quantity and finish, and the quote comes back with tolerance and process notes, not just a price.