Screw Holes Are Where Your IP67 Enclosure Leaks — Three Ways to Seal a Screw Boss

Most waterproof enclosures don't fail at the gasket — they fail at the screw boss. Compare the three standard screw-to-gasket layouts, what each costs in BOM, assembly time, and mold complexity, and how to pick one before you cut steel.

METAL PARTS

Peakingtech

8/13/20266 min read

the three layouts, plan view (the spine of the article).
the three layouts, plan view (the spine of the article).

When a sealed product comes back from IP testing with water inside, the instinct is to blame the gasket. Almost always, the gasket is fine. The leak path runs through a screw boss.

This is not bad luck. It is the direct consequence of asking a single screw to do two jobs that pull against each other. The screw has to generate enough clamping preload to compress the main gasket to its design squeeze — and at the same time, the hole it passes through must never become a channel from the outside world into the cavity. Every waterproof enclosure resolves that tension one of three ways, and the choice is usually made early, quietly, by whoever draws the housing.

It is worth understanding before it is made for you, because the three options differ enormously in tooling difficulty, per-unit assembly labour, and whether your line can be automated later.

Option 1: Route the gasket around the boss, keeping the screw inside the seal

The main gasket path detours outward around each boss, so the screw hole sits inside the sealed loop. There is no separate seal at the screw — the main gasket is doing all the work.

Why teams choose it. It is the lowest part count of the three. One gasket, no per-screw hardware, nothing small for an operator to forget. Assembly is a single ring, dropped in once. And because the boss is inside the loop, the clamping force from each screw lands directly on the sealing face, so preload converts efficiently into gasket compression.

Where it hurts. The gasket path now has a bulge and two tight corners at every boss. Sharp corners are the hardest feature to mould in a silicone gasket — that is where you get short shots, thin sections, and knit lines. They are also where the housing groove is hardest to fill and hardest to hold to tolerance. The result is a compression ratio that swings around the corners: correct along the straights, too high on the outside of a bend, too low on the inside. At IPX4 that is survivable. At IP68 with pressure and thermal cycling, those corners are your failure population.

Best fit: moderate IP targets, a housing with enough footprint that the detours can be generous radii rather than sharp elbows, and a programme where BOM cost matters more than a few points of first-pass yield.

Option 2: Put the screw inside the cavity and seal it separately

Here the bosses live entirely inside the sealed volume, and the main gasket runs a clean, smooth loop with no detours at all. Each screw gets its own seal — a bonded washer or a small O-ring under the head — so the screw hole is closed independently of the main seal.

Why teams choose it. The main gasket becomes the easy part. A smooth path moulds at high yield, the groove is simple to machine and simple to polish, and compression ratio stays constant all the way around. Just as importantly, the two sealing systems are decoupled: over-torquing one screw disturbs the local gasket compression far less, because the primary seal is no longer wrapped around the boss.

Where it hurts. You have added a part and an operation per screw. Eight screws means eight small O-rings to buy, kit, and fit — and each one is a chance to omit, twist, roll, or pinch. Small elastomer parts are exactly what manual assembly gets wrong at 3 a.m. on the second shift. Worse for scaling, an automatic screwdriver cannot easily present a screw with a loose O-ring on it; you are designing yourself into hand assembly unless you switch to pre-assembled bonded (SEMS) washers, which cost more per piece but restore automation.

Best fit: high IP targets where the main seal must be beyond argument, products with internal room to spare, and lines where labour is available and volumes are moderate.

Option 3: Put the screw outside the loop and bend the gasket inward

The bosses move outboard, onto the flange or the housing edge, entirely outside the gasket loop. The gasket bends inward to route around each one. Water that gets into a screw hole simply sits there — it has no path across the seal into the cavity.

Why teams choose it. This is the space-efficient option. Bosses on the outside of the seal do not eat internal volume, which matters enormously when the PCB, battery, and antenna are already fighting each other. There is no per-screw seal, so no small parts and no automation penalty. For a product being pushed toward miniaturisation, this is often the only layout that closes.

Where it hurts. The gasket contour is now the most complex of the three — multiple inward bends, many corners, and a moulding job that is genuinely difficult to hold in tolerance. Compression control at those corners is the hardest of the three options, with local over-compression and under-compression sitting side by side. And you have permanently exposed the fasteners: brass heat-set inserts, insert-moulded studs, and screw heads now live in the wet zone. Over a product's life they corrode, and corrosion debris can migrate to the sealing edge. Material selection stops being a detail — stainless fasteners, compatible insert material, and a sealed or recessed screw head become mandatory rather than nice to have.

Best fit: compact products, automated or semi-automated assembly, and teams with a gasket supplier who can genuinely hold a complex profile.

Choosing between them

The ranking flips depending on which constraint is binding. Ask, in this order: what IP rating are you actually contractually committed to; is internal volume tight; will this line ever be automated; and can your gasket supplier hold a complex profile at the yield you need. The answer usually falls out after the second question.

An assembly cost check worth running

Option 2's extra parts look trivial until you multiply. Take an 8-screw enclosure: eight O-rings at your piece price, plus the seconds an operator needs to fit each one, plus the kitting and line-side stocking of one more part number, plus the scrap from ring omission caught at test. Run that against your own labour rate and volume before accepting "it's only an O-ring." On a 20,000-unit programme it is frequently the difference between the three options — and it is a number your CM can produce for you in an afternoon.

Five rules that apply whichever layout you pick

  1. Match preload to target compression, not to a torque that "feels tight." For a static face seal, target roughly 20–25% squeeze, with a workable band of about 15–30%. Below that the seal is not energised; above it you accelerate compression set and the seal relaxes early in the field.

  2. Do not overfill the groove. The gasket needs somewhere to go when it is compressed. Industry practice keeps gland fill in the 60–85% range and treats 90% as a hard ceiling — beyond that the groove hydraulically locks, the ring cannot compress further, and thin-walled housings can be damaged during assembly.

  3. Rib the housing around every boss. Tightening a screw deflects the housing, and housing deflection changes gasket compression directly. If the flange bows between bosses, no sealing layout will save you. Ribs and boss spacing are part of the seal design, not cosmetic stiffening.

  4. Radius every corner in the gasket path. Corners are where moulding defects and compression variance both concentrate. Generous radii cost you a little board area and buy you first-pass yield.

  5. Design the fastener for its environment. If the screw lives outside the seal, specify stainless, pick an insert material that will not couple galvanically with it, and recess or cap the head.

Validate before you cut steel

None of the above is a decision you make once on a CAD screen. Build soft-tool or 3D-printed housings with the real gasket profile, assemble at production torque, and test: IPX7 immersion per IEC 60529, thermal cycling followed by re-test, drop then re-test, and a compression-set check after ageing. Section the parts afterwards and measure actual squeeze at the corners, not just on the straights — that is where the design margin either exists or does not.

There is no universally correct answer here, and any vendor who gives you one without asking about your IP target, your internal volume, and your assembly line is guessing. The right layout is the one that fits your space, your rating, your tooling budget, and — the constraint most often forgotten — the process capability of the factory that will actually build it.

FAQ

Where do waterproof enclosures usually leak? At the screw bosses, not along the gasket run. The boss is where the gasket path is interrupted, where compression is least uniform, and where housing deflection under fastener load is highest.

What compression ratio should a waterproof gasket have? For a static face seal, aim for about 20–25% squeeze. Roughly 15–30% is the usable band. Too little and the seal never energises; too much causes compression set and early field failure.

How full should the gasket groove be? Keep gland fill between about 60% and 85%, and never exceed 90%. The elastomer must have room to deform sideways when compressed, and to expand with heat.

Do I need a separate O-ring on every screw? Only if the screws sit inside the sealed cavity and pass through to the outside. If the main gasket encircles the bosses, or the bosses sit outside the sealed loop entirely, the screws need no individual seal.

Which layout is best for IP68? A smooth, uninterrupted main gasket path is the most reliable at IP68 — which favours putting the screws inside the cavity with their own seals. The outside-the-loop layout can also reach IP68, but demands much tighter control of gasket moulding and compression at the corners.

Does screw torque affect waterproofing? Directly. Torque sets preload, preload sets gasket compression, and housing deflection under that preload changes compression again. Torque should be derived from the target squeeze and verified on real parts, not inherited from a similar-looking product.

Peakingtech provides NPI and contract manufacturing for hardware startups and product brands, including DFM review of sealing and enclosure design, soft-tool pilot builds, and IP validation before production tooling is committed. Talk to us about your enclosure.