Short answer: size the groove to the cord, then the ring to the path. For a 3 mm [0.12 in] cord, mill the groove 2.25 mm [0.089 in] deep and 4.0 mm [0.16 in] wide with 0.4 mm inside corners. That squeezes the ring 25% when the lid closes and fills the gland to 78.5% — inside the 20-30% squeeze band and the 85% fill ceiling the O-ring handbooks set for face seals. Cut or order the ring 1-2% short of the groove's centreline path, and round the path corners to at least twice the cord diameter. The worked example below sizes the groove for a 120 x 80 mm screw-down lid; the steps then take it to a case you can sink.
The three numbers: squeeze, fill and stretch
An O-ring seals because the closed gland squeezes its cross-section, and the compound pushes back like a spring. Three numbers set that up, and the Parker O-Ring Handbook tabulates all three — it is the reference to keep on the bench.
Squeeze is the flattening of the cord: cord diameter minus gland depth, over cord diameter. Face seals run 20-30% squeeze, the top of the general 10-30% band, because a flat cover presses the ring evenly over its whole face. Design at 25%: force enough to seal, without stressing the housing.
Fill is the share of the groove the ring occupies: ring cross-section area over groove cross-section area. Handbooks hold fill under an 85% ceiling to leave room for cord tolerance and thermal swell, and work near 75% so the ring cannot shift in the groove. Depth sets squeeze; width sets fill.
Stretch is how much longer the groove path is than the ring's free circumference. Hold it at 1-2%: enough to seat the ring while you fit the lid, little enough to leave the cord's cross-section alone. Each 2% of stretch thins the cord about 1%, and it trims your squeeze with it. How squeeze turns into sealing force is the subject of O-ring squeeze for a sealed enclosure face.
Two rules carry to any cord: depth = 0.75 x cord, width = 1.33 x cord. They land every common cord at 25% squeeze and 78-79% fill, with the groove's inside corners rounded at 0.3-0.5 mm as the handbook gland tables do:
| Cord | Groove depth | Groove width | Inside corner radius |
|---|---|---|---|
| 2 mm | 1.50 mm | 2.66 mm | 0.3 mm |
| 3 mm | 2.25 mm | 4.00 mm | 0.4 mm |
| 4 mm | 3.00 mm | 5.32 mm | 0.5 mm |
| 1/8 in (3.53 mm) | 2.65 mm | 4.70 mm | 0.4 mm |
A four-lobed X-ring drops into the same gland numbers and resists rolling; on a static face it is a refinement, and a plain O-ring does the job.
Sizing the groove for a 120 x 80 mm lid with a 3 mm cord
Inputs. A machined lid 120 x 80 mm over the groove centreline, path corner radius 6 mm — twice the cord, so the bend stays gentle. Cord 3 mm [0.12 in], 70 Shore A nitrile at the standard hardness; nitrile suits fresh water and spans -30 to 120 C in the handbook tables. Squeeze target 25%, the middle of the face-seal band; fill ceiling 85% with 75% as the working figure.
Depth, from squeeze. Depth = cord x (1 - 0.25) = 3.00 x 0.75 = 2.25 mm [0.089 in]. Check: (3.00 - 2.25) / 3.00 = 25%.
Width, from fill. Width = 1.33 x 3.00 = 4.0 mm [0.16 in]. Ring area = π/4 x 3.00² = 7.07 mm². Groove area = 4.0 x 2.25 = 9.0 mm². Fill = 7.07 / 9.0 = 78.5%; the inside-corner radii shave both areas roughly equally, so the figure stands. Past 85% fill, widen the groove in 0.2 mm steps — each step drops fill about 4 points.
Ring length, from the path. Straight runs: 2 x (120 + 80) = 400 mm. The four corners replace 8 x 6 = 48 mm of straight with 2π x 6 = 37.7 mm of arc, so the path is 400 - 48 + 37.7 = 389.7 mm — call it 390 mm. At 1-2% stretch, the free circumference is 390 / 1.02 = 382 mm to 390 / 1.01 = 386 mm. Cut cord stock to 382-386 mm; for a one-off, splice it to length — bonding spliced O-ring cord into a closed ring shows the joint — and keep the splice on a straight run.
Result. Groove 4.0 mm wide x 2.25 mm deep, inside corners 0.4 mm, path corners 6 mm, 0.3 mm chamfer on the groove lip, cord at 382-386 mm. Depth is the sensitive input: each 0.05 mm of depth moves squeeze 1.7 points. With ±0.05 mm held on the depth and ±0.08 mm allowed on the cord — swap in your shop's numbers and the ring datasheet — squeeze lands between 21% and 29%, inside the band.
Cutting the groove
What you need: an endmill at the groove width (one 4.0 mm pass for the example), a corner-rounding tool or abrasive cone, a fine stone, a depth gauge, the cord and its datasheet, and a grease the compound tolerates. Petroleum grease suits nitrile; EPDM needs silicone grease.
- Face the land flat before cutting. Depth runs from this finished face to the groove floor, so any rework or coating afterwards costs squeeze.
- Mill the path full width in one pass, rough to depth, then take a 0.2 mm finish pass so the walls stand straight and clean.
- Gauge the depth at four points, two of them in corners: 2.25 ±0.05 mm for the example. Corners cut shallow while the endmill load settles.
- Round the groove's four inside corners to 0.4 mm and break the groove lip with a 0.3 mm chamfer. A sharp lip shaves the cord every time the lid closes.
- Bring the surfaces to the handbook's static-seal figures: 0.4 µm [16 µin] Ra on the land, 0.8 µm [32 µin] on the groove walls, 1.6 µm [63 µin] on the floor. As-milled faces sit at 1.6-3.2 µm, so stone or re-face the land.
- Clear the chips and wipe the groove dry. A chip under the cord is a leak path you can point at later.
- Lay the greased ring into the groove. It should stand proud of the floor and hold its shape through the corners without forcing.
- Close the lid and screw down in a criss-cross pattern, a turn at a time, so the land seats flat before any screw pulls tight.
Checks before you close the case
Gauge twice: depth at 2.25 ±0.05 mm, and a straightedge across the land with a 0.05 mm feeler that finds no gap. Then prove the seal with water: leave the closed case 1 m under water for 30 minutes — the depth and time IEC 60529 uses for its second digit 7 — with a dry paper towel inside. A dry towel, plus no fogging when you run cold water over the lid, means the face holds. Immersion even helps a face seal: depth pushes the lid onto the ring, so a joint that passes at the bench grips harder under water.
Two things live outside the groove. A case assembled in humid air fogs on the first cold night even with a perfect seal, so assemble dry and drop in desiccant; the desiccant guide for sealed enclosures sizes it. And the face is rarely where these cases leak — the cord entry is. A molded plug through the wall is its own sealing problem: getting a power cord with a molded plug through an enclosure knockout covers the cord, and the cutouts, glands and windows guide covers entries generally.
Where rectangular grooves go wrong
- Path corners tighter than one cord diameter. The cord kinks against the inside of the bend and the corners weep first; hold the path radius at twice the cord, 6 mm on the example.
- Depth cut from the wrong datum. Coatings applied after machining join the gland: 0.05 mm of hardcoat on both sealing faces lifts a 25% design to about 27%. Mask the sealing faces or cut the depth last.
- A cover held off its seat. A 0.2 mm standoff gap turns 25% squeeze into 18%, and the ring floats. Sight for daylight around the land after the first torque pass.
- Stretch past 5%. The cord thins about 2.5% and squeeze slides to about 23%; cut the ring to the 1-2% figures for your path instead of forcing a stock ring around it.
- A land left with cutter swirl. Marks at 1.6 µm Ra and coarser channel water straight across the face; stone the land to 0.4 µm.
- A ring twisted in the groove. Feed it in greased and flat; a quarter-turn of twist reads as a leak at one corner.
O-ring groove, gasket or raised lip: the decision
Pick the seal by how the case lives, and put the numbers on your drawing:
| Case | Seal | Numbers to hold |
|---|---|---|
| Screwed metal lid, occasional immersion to about 1 m | O-ring groove as worked above | 25% squeeze, 78.5% fill, 1-2% stretch, land at 0.4 µm Ra |
| Rain and splash, lid opened weekly for service | Compression gasket in a tongue-and-groove face | 20-30% gasket compression, held by the tongue — see tongue-and-groove gasket design for a sealed enclosure |
| Flat faces you reopen often | Flat gasket against a raised lip | the lip caps compression at the designed squeeze — see raised lip or flat face gasket in a sealed enclosure |
| A product box rather than a machined one-off | Molded-gasket stainless enclosure | rated NEMA 4X / IP66 by the manufacturer; not UL listed — start at custom stainless enclosures |
For the underwater case, the groove above is the whole answer: 2.25 mm deep, 4.0 mm wide, 6 mm path corners, ring 382-386 mm around a 390 mm path, proven at 1 m for 30 min before its first real dunking.
Catalog dimensions to compare
Compare these catalog dimensions with your component layout. Open each enclosure for its rating, material and usable space, then check those specifications against your application.
| Enclosure | Material | Outside H x W x D / display | Rating |
|---|---|---|---|
| 12 × 10 × 6 in | 304 stainless | 12 x 10 x 6 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
| 16 × 14 × 8 in | 304 stainless | 16 x 14 x 8 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
| 20 × 16 × 8 in | 304 stainless | 20 x 16 x 8 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
Questions
Does the O-ring circumference have to match the groove length exactly?
Match the groove's centreline path, then buy the ring 1-2% shorter: a 390 mm path takes a ring of 382-386 mm free circumference, which stretches into place and stays put while you fit the lid. At zero stretch the ring can shift during assembly; past about 5% it thins the cord enough to pull squeeze out of the 20-30% band.
What happens above 30% squeeze on a face seal?
Sealing force keeps rising but the failure risk rises faster: above 30% squeeze the ring works hard against the housing, the extra load bows thin lids between the screws, and the cord can pinch or shear at the groove lip when the cover closes. Design at 25%; stay inside 20-30%; every 0.05 mm of extra depth removes about 1.7 points.
Can I build the ring from cord stock instead of a molded ring?
Yes for a one-off: cut the cord to the worked 382-386 mm and bond the ends with the cord maker's splice kit. Let the joint cure fully before you grease the ring — an uncured butt joint lets go under the first clamp load. The splice stays the weak point, so a molded ring wins for anything you assemble more than a few times.
Is a self-machined groove rated like a catalog enclosure?
A rating belongs to a tested product: NEMA 250 and IEC 60529 describe enclosures that passed their published tests, and your milled case starts unrated. The 1 m / 30 min immersion check shows the face holds at the IPX7 depth and time, and water pressure on the lid only adds clamp load. A catalog stainless box carries a tested rating instead; a groove you mill yourself is a design, not a rating.
