The Seal That Actually Seals
The Seal That Actually Seals
Your enclosure passed IP65 on the bench and leaks in the field, because the gasket took a set after two years of temperature cycling and the joint designed at 30% compression is now working at 15%. The seal is a set of numbers, not a material.
The enclosure passed IP65 on the bench. The test report says so, dust-tight, water jets from any direction, three minutes, 12.5 litres a minute at 30 kPa. The certificate is in the file. The product shipped.
Two years later, the field units are leaking. Not all of them. Not on day one.
The ones that have been outside for two summers and two winters, the ones that have been hot in the sun and cold at night, two hundred times over, are letting water in around the lid. The gasket is still there. It looks fine.
It is just shorter than it was, and it does not push back the way it did, and the joint that was designed at 30% compression is now working at 15%, and 15% is the floor, the point where the seal stops filling the microscopic gaps in the lid and the water finds its way in through channels too small to see.
The seal did not fail because the material was wrong. It failed because the seal is not a material. The seal is a set of numbers, the compression, the groove depth, the groove fill, the fastener spacing, the corner radius, the pressure the box breathes, and one of those numbers drifted, slowly, over two years, until the joint crossed the floor.
The bench test could not see it, because the bench test is a snapshot, and the failure is a drift. The seal that actually seals is the one whose numbers still hold after two years of temperature cycling, and this article is the arithmetic for checking them.
The simple idea first
Think of a gasket as a spring. Not a metal spring, a rubber one, a ring of rubber that sits in a groove, and when the lid closes, the lid squeezes it, and the squeezed rubber pushes back, and that push-back is the seal. The water cannot get past the joint because the rubber is pressing against both faces, filling every microscopic gap, with force.
And like any spring, the rubber has a working range. Squeeze it too little, and it does not push back hard enough to fill the gaps, the water walks in. Squeeze it too much, and it takes a permanent set, it stays flat, it stops pushing back, and the joint opens.
The working range is narrow: roughly 25 to 35 per cent compression for a door gasket, 15 to 30 per cent for an O-ring. The whole job of gasket design is staying inside that range, on day one, and after two years of temperature cycling, and after the lid has been opened 25 times for battery service.
And the range is set by the groove, not by feel. The groove depth is what sets the compression, the gasket's free height minus the groove depth, divided by the free height. If the groove is 0.1 mm too deep, the compression falls.
If the groove is powder-coated, the coating eats 0.06 to 0.08 mm of depth, and the compression rises. If the lid bows between the screws, the compression is 30% at the screws and 15% in the middle. The groove is the design.
The gasket is just the rubber that fills it.
The rating belongs to the assembly, not the gasket
IP65 is not a property of a panel or a gasket. It is a property of a finished assembly, and it is set by the worst joint on that assembly. (Sanjun) The rating belongs to the whole enclosure, so the lowest-performing joint sets the result. A door perimeter sealed to a standard you would be proud of earns nothing if the top-left cable gland is a plain hole with a grommet that was never specified. (Sanjun)
And IP65 covers dust-tightness and water jets only. It does not cover immersion, washdown or corrosion. (Sanjun) The second digit is the water digit, and each step up is a different test, not a harder version of the same test, Article #7's point, carried forward: IPX5 is 12.5 litres a minute at about 30 kPa, IPX6 is 100 litres a minute at about 100 kPa, eight times the volume, three and a third times the pressure, and IPX7 is immersion, 1 metre for 30 minutes, and IP69K is 80 °C at 80 to 100 bar.
Above IPX6 the water digits are separate tests, not a ladder. The enclosure that passes IP65 has passed a jet test. It has not passed an immersion test, and it has not passed a two-year temperature-cycling test, because neither of those is in the rating.
The five joints that fail first are the ones to check, in the order an enclosure fails them. (Sanjun) The door-to-frame perimeter, the gasket is only as good as what presses on it, and hinge and latch geometry decide whether that pressure is even. The cable glands and connector cutouts, one missed grommet in a bank of twenty is the whole rating gone.
The vent panels and filter fans, air and water want to travel in opposite directions through the same opening. The panel-to-panel seams on large cabinets, flatness and weld quality decide this one, not the gasket. And fastener spacing on removable covers, too few screws means uneven compression, and uneven compression is a gap.
The fifth is the one most often missed, and it is also the cheapest to get right at drawing stage: adding screws later means a new tool, a new torque instruction and a new assembly step for a problem that was visible on a CAD screen. (Sanjun)
The hub has five joints of its own: the lid-to-base perimeter, the cable gland where the sensor cable enters, the vent where the Gore membrane sits, the battery door, and the eight M4 screws that hold the lid. The lid gasket is the one the buyer is looking at. The cable gland is the one that is probably leaking.
The article's demo checks the lid gasket, because that is the joint the buyer named, and it names the other four, because the rating belongs to the assembly, and the worst joint sets the result.
The window: 25 to 35 per cent, set by the groove
Door gaskets need a working compression of roughly 25 to 35 per cent, set by groove depth rather than by feel. (Sanjun) A door seal is only as good as the groove it sits in. Groove depth, groove width and corner radius together determine how much the seal is squeezed when the door shuts, and that number is what makes or breaks the joint.
For closed-cell sponge we work to roughly 25 to 35 per cent compression of free height. Compress it more and the gasket takes a permanent set, after which it stops pushing back and the joint opens; compress it less and the first thermal cycle is enough to lift it off. (Sanjun)
The window has hard edges, and the edges have names. Under-compression, below 15%, fails to fill microscopic irregularities in the door panel, creating capillary siphon channels. Over-compression, above 50%, exceeds the material's yield limit, leading to permanent set and loss of long-term rebound capability.
(Hinges Manufacturers, citing ASTM D395 / ISO 815) Below 15%, the water does not need a gap. It needs a capillary, a microscopic channel between the gasket and the lid, narrow enough that surface tension pulls the water along it, the way water climbs a thin tube. The gasket can look sealed and still leak, because the leak is at the scale of the surface finish, not the scale of the eye.
And the window is not the same for every seal. O-rings sit in a machined or moulded groove and seal through controlled compression, typically 15 to 30 percent. (ATI Propel) The standard confirms it: ISO 3601-2 gives an initial squeeze of 15 to 30% for static applications, and 6 to 20% for dynamic ones.
(Engineers Edge, citing ISO 3601-2) A static seal, a lid that closes and stays closed, tolerates the upper end. A dynamic seal, a shaft that moves, lives near the bottom, because the squeeze is friction, and friction is heat, and heat is wear. The hub's lid is a static seal.
The hub's battery door, opened 25 times over a 10-year life, is a static seal that is opened often enough to matter, and Article #49's rule applies: inserts are needed when a joint is opened or re-fastened more than 5 to 10 times, and the hub's lid is opened 25 times.
The other numbers that hold the seal are just as specific. The groove fill, the gasket's cross-section over the groove's cross-section, sits at 70 to 85%. Below 70% the gasket walks; above 85% it extrudes and lifts the panel.
(Xingyuan) The O-ring version of the same rule: the ring should occupy about 75 to 90 percent of the groove volume, leaving room for thermal expansion and fluid swell. A completely filled groove hydraulically locks and can split the housing. (Projects House) The seat finish: machine gasket seats to 1.6 µm Ra or better, and check flatness after machining, a seat that is smooth but dished will still leak.
(xkyhardware) The corners: below 1.5× cord diameter the gasket wrinkles at a 90° bend, and a wrinkle always leaks. Use moulded 45° corner pieces and keep splices away from latches and the bottom centre. (Xingyuan) And the butt joint: butt joints in a gasket run are planned leaks.
They are sometimes unavoidable, and when they are, they belong at the top of the opening, sloped so water runs out rather than in, with the cut ends sealed. A butt joint at the bottom of a door is a gutter pointed at your electronics. (Sanjun)
The groove tolerance is the quietest killer in the list. Groove depth is what sets compression. A groove 0.1 mm too deep removes roughly a third of the compression on a 3 mm cord.
(xkyhardware) And coating eats compression: 60 to 80 µm of powder inside the groove removes 0.06 to 0.08 mm of depth. (Xingyuan) The groove is machined to a tolerance, and the tolerance is the compression, a 0.1 mm error on a 2.1 mm groove is a 5% error in depth, and a 5% error in depth is a third of the compression gone, because the compression is the difference between two numbers that are close together.
The drawing has to say the groove depth, and the groove depth has to be measured, because the groove depth is the seal.
Compression set: the gasket that gets shorter
Compression set is the measure of how much a specimen fails to recover after being held compressed. It matters because a gasket that has taken a set is shorter than it was, so it pushes back with less force, so the joint that was designed at 30 per cent is now working at 15. (Sanjun, paraphrased, the pipeline's wording)
The test is ISO 815-1:2019, "Rubber, vulcanized or thermoplastic, Determination of compression set, Part 1: At ambient or elevated temperatures", which specifies methods for the determination of the compression set characteristics of vulcanized and thermoplastic rubbers at ambient or elevated temperatures, for rubbers of hardness within the range 10 IRHD to 95 IRHD, after prolonged compression at constant strain, normally 25%. (ISO 815-1:2019) When rubber is held under compression, physical or chemical changes that prevent the rubber returning to its original dimensions after release of the deforming force can occur.
The result is a set, the magnitude of which depends on the time and temperature of compression as well as on the time, temperature, and conditions of recovery. At elevated temperatures, chemical changes become increasingly more important and lead to a permanent set. (ISO 815-1:2019)
The formula is simple, and it is the whole article in one line:
where t₀ is the original specimen thickness, tᵣ is the recovered thickness after the specified recovery period, and tₛ is the spacer thickness during the test. (Elastotech) A specimen originally 10.0 mm thick, compressed to 7.5 mm, that recovers to 9.5 mm has a compression set of (10.0 − 9.5) / (10.0 − 7.5) × 100 = 20%. (Elastotech) Twenty per cent of the imposed compression never came back. The specimen is 0.5 mm shorter than it was, and it will never be 10.0 mm again.
The reference standard for the test is ASTM D395 or ISO 815. (Hinges Manufacturers) And the test conditions matter: a standard number alone is not enough to compare two percentages fairly, ASTM D395 includes compression-set test methods using constant force or constant deflection, and ISO 815-1 addresses compression set at ambient or elevated temperatures, and the set depends on the time and temperature of compression and on the time, temperature, and conditions of recovery.
(Elastotech; ISO 815-1:2019) A gasket that shows 20% set after 24 hours at 70 °C is not the same as a gasket that shows 20% set after 2 years at 25 °C, and the field gasket is the second one, and the datasheet is usually the first.
The material decides how fast the set grows. Closed-cell silicone sponge, 10 to 25 Shore A, is the default, best rebound, UV and ozone stable. (Xingyuan) Solid silicone or EPDM, 50 to 70 Shore A, is for high differentials and stiff doors, EPDM is cheaper but hardens above about 90 °C.
PU foam is convenient for irregular shapes, but takes a large compression set and needs replacing within two years. (Xingyuan) Article #7's field lives, carried forward: EPDM 8 to 10 years (−40 °C to 120 °C), silicone 10 to 15 years (−60 °C to 200 °C). And Article #7's loss figure: gaskets lose about 20% of sealing after 2 years of temperature cycling.
The PU foam gasket is the one that takes the set the fastest, and it is the cheapest one, and it is the one in the hub, because the hub's gasket was chosen on price, and the price was right, and the set is the cost.
The fasteners: the spacing that holds the compression even
Screw-compressed seams use a continuous gasket pressed by evenly spaced fasteners; the rule of thumb is fastener spacing no greater than 40 millimetres to prevent gasket bulging between screws. (ATI Propel) Too few screws means uneven compression, and uneven compression is a gap. (Sanjun) The gasket bulges between the screws because the lid is a beam, and the beam bends between its supports, and the bend lifts the lid off the gasket in the middle of the span, so the compression is 30% at the screws and 15% in the middle, and the joint leaks in the middle, where nobody looks.
The numbers from the sheet-metal guides agree with the 40 mm rule and tighten it for stiffer demands: fastener pitch ≤ 75 mm for IP65 on sheet 1.5 mm or thicker, ≤ 50 mm for IP66 on sheet 2.0 mm or thicker, ≤ 120 mm for IP54. (JS Precision) And the uniformity target: the uniformity of compression throughout the entire circumference is within 10%. (JS Precision) The compression has to be even to within a tenth, not because the gasket needs it, but because the weakest point sets the rating, and the weakest point is the middle of the longest span.
The hub's lid is held by 8 × M4 screws around a 320 mm perimeter, 40 mm spacing, exactly at ATI Propel's limit. The spacing is right, and it is right by exactly nothing, one screw fewer, and the spacing is 45.7 mm, and the lid bulges, and the joint leaks in the middle of the span. The demo works the number.
And the screws are not the only way to hold the compression. Compression latches differ fundamentally from standard cam latches by their ability to provide a 3 mm to 6 mm axial stroke after closing. This stroke generates a controlled preload, forcing the gasket into its effective compression range (20% to 40%), thereby ensuring compliance with IP65/66 and NEMA 4/4X standards.
(Hinges Manufacturers) The core value of a compression latch lies in separating the "positioning" and "compressing" actions, the latch first positions the door, and then a second motion pulls it axially against the frame, by 3 to 6 mm, actively eliminating the door gap. (Hinges Manufacturers) The latch is the hard stop and the preload in one part: it closes the door to the same place every time, and it squeezes the gasket by the same amount every time.
The screw does the same thing, if it is torqued the same way every time, and the screw is torqued by a person, and the person is not a latch.
The vent: the sealed box that breathes
A sealed enclosure is not a static system. When the product warms up in sunlight or cools down in a refrigerator, the air inside expands and contracts. The pressure difference can be large enough to push water past a seal, distort a membrane, or pop a battery door. This is why almost every serious waterproof product includes a small breathable vent, typically a Gore-Tex style ePTFE membrane, that lets air pass but blocks liquid water. (ATI Propel)
The physics is the ideal gas law, and the numbers are large. A sealed 600 × 400 × 250 mm box rising 20 K carries about 6.7 kPa of pressure difference (ΔP = P·ΔT/T), and the reversed differential at night pulls the gasket lip inward. (Xingyuan) The better the seal, the larger the swing, and the harder it pulls water through its weakest joint.
(Xingyuan) The seal that is good enough to hold the pressure is also the seal that has to hold the pressure, and the pressure is not in the IP test, because the IP test is at room temperature, and the field is not.
Gore's case study shows it directly. Two identical enclosures, each with a heat source, a gasket-sealed cover, and a balloon to show the pressure. Both are exposed to a temperature increase from 20 °C to 50 °C in 90 seconds, in a control volume of 4,000 cm³.
When heat is applied, the resulting increase in internal pressure within the unvented enclosure causes the balloon to inflate. In contrast, the vented enclosure maintains stable internal conditions because the Gore Protective Vent enables air exchange, allowing pressure to equalize with the external environment. Then the same enclosures are exposed to a 90-second rapid cooling cycle from 20 °C to −5 °C.
The decreased air pressure in the unvented enclosure creates a significant vacuum, causing the balloon to deflate. In a real-world situation, contaminants such as liquid, dirt and particles can be drawn into a device, causing failure to internal components. (Gore, "The Impact of Temperature Cycling on Sealed Enclosures")
The vent is a membrane, not a hole. A hydrophobic vent from Gore has a microporous structure with pores too small for water molecules to pass through, water has high surface tension, so it is more likely to sit atop a vent membrane than to permeate its pores. An effective hydrophobic membrane allows enclosures to breathe and maintain a stable internal pressure, while preventing water in various forms from entering the enclosure.
(Gore) The Gore Protective Vents Adhesive Series, VE8, VE7, VE9, have typical airflows at 70 mbar of 3,300, 290, and 1,150 ml/min/cm², and over 200 million installations worldwide. (Gore) And the vent has a rating of its own: the POV/M12x1 screw-in vent is IP65 for water jets, IP67 for 1 metre submersion for 30 minutes, and IP69K for high-pressure spray. (Gore, via Owlogic) The vent is not a hole in the seal.
It is a seal that lets air through and keeps water out, and it is the cheapest fix for a product that mysteriously fogs or drips inside. (Projects House)
And the vent's position matters. A vent on the bottom of a wearable, recessed under a small overhang, is far more reliable than a vent on a flat top surface. (ATI Propel) The vent on the top is a puddle. The vent on the bottom, under an overhang, is dry. The hub's vent, if it has one, belongs on the bottom of the base, under the lid's overhang, not on the flat top of the lid.
The seam: three ways to close a box, one of them permanent
A waterproof housing usually consists of two halves that must seal against each other for the entire product life. Three approaches dominate. (ATI Propel) The choice between them is the choice between a box that opens and a box that does not, and the hub has to open, 25 times over a 10-year life, for battery service and calibration, so the choice is already made. But the buyer should know what the other two cost, because the choice is the reason the seal has to be designed at all.
Screw-compressed seams use a continuous gasket pressed by evenly spaced fasteners, the hub's approach, fastener spacing no greater than 40 mm. (ATI Propel) The seam can be opened and closed, the gasket can be replaced, and the compression can be checked. The cost is the fasteners, the torque, and the discipline: every screw has to be torqued the same way, every time, and the gasket has to be replaced when it takes a set, and the groove has to be cleaned when the lid opens. The screw-compressed seam is the only one of the three that can be serviced, and it is the only one that can be serviced wrong.
Snap-fit seams with integrated seals are common for low-cost consumer products but rarely exceed IPX4 reliably. (ATI Propel) IPX4 is splashes from any direction, rain, not jets. The snap-fit holds the halves together with clips, and the clips cannot hold the compression even, because the clips are at the corners and the gasket bulges between them, and the clips relax with temperature, and the seal is the clip's spring, and the clip's spring is plastic, and plastic creeps. The snap-fit seam is cheap, and it is the reason the product is cheap, and the product leaks in the rain.
Ultrasonic welding fuses the two halves into a single part and is the most reliable seal for sealed-for-life products like wearables and medical disposables. (ATI Propel) There is no gasket, because there is no joint, the two halves are one part, and the water has no path. The cost is that the box never opens again. The battery cannot be replaced. The board cannot be recalibrated. Article #7's site rule applies: "Made to be kept, instruments designed to be recalibrated, not replaced." The hub is an instrument, and it is recalibrated, and it opens 25 times, so the weld is not an option, and the gasket is the seal, and the gasket has to be designed.
The seal that leaks on the hub board
Example Labs is a made-up company we use for worked examples.
The Example Labs field sensor hub: 100 × 60 × 40 mm, aluminium base, polycarbonate lid, 8 × M4 screws, opened 25 times over a 10-year life (labelled assumption, from Article #49). The lid gasket is a closed-cell silicone sponge cord, free height 3.0 mm (labelled assumption, a 3 mm cord is the standard example size, and the groove-depth rule is stated for a 3 mm cord). The groove is 2.1 mm deep and 4.5 mm wide.
The enclosure passed IP65 on the bench. Two years later, the field units are leaking around the lid.
Step 1, the design. The compression is (3.0 − 2.1) / 3.0 = 30%. The squeeze is 0.9 mm. The groove depth for 30% compression on a 3.0 mm cord is 2.1 mm, inside Sanjun's 25 to 35% window, which is a groove depth of 1.95 to 2.25 mm. The groove fill: the cord's cross-section is π × 1.5² = 7.07 mm², the groove's is 2.1 × 4.5 = 9.45 mm², so the fill is 7.07 / 9.45 = 74.8%, inside the 70 to 85% window, and inside the O-ring rule of 75 to 90%. The design is right. The bench test passed. The numbers hold, on day one.
Step 2, the set. After two years of temperature cycling, the gasket takes a compression set. Take TEMCO's worked example: a specimen originally 10.0 mm thick, compressed to 7.5 mm, that recovers to 9.0 mm has a compression set of (10.0 − 9.0) / (10.0 − 7.5) × 100 = 40%. The hub's gasket was compressed by 0.9 mm. A 40% set means 0.36 mm never comes back. The gasket's free height is now 3.0 − 0.36 = 2.64 mm. The working compression is (2.64 − 2.1) / 2.64 = 20.5%. The joint that was designed at 30% is now working at 20.5%, still above the 15% floor, but the margin is gone. The bench test is two years old, and the gasket is 0.36 mm shorter, and the seal is one thermal cycle from the floor.
Step 3, the tolerance. The groove is 0.1 mm too deep, 2.2 mm instead of 2.1 mm, a machining tolerance the drawing allowed. The working compression with the set gasket is (2.64 − 2.2) / 2.64 = 16.7%. The joint is 1.7 points above the 15% floor, one thermal cycle from a capillary leak. And the buyer's 15% is the same mechanism one step further: for the working compression to be 15% with a 2.1 mm groove, the gasket's free height must be 2.1 / 0.85 = 2.47 mm, a loss of 0.53 mm, a compression set of 0.53 / 0.9 = 58.8%, about 59%. That is the kind of set PU foam takes, and PU foam "takes a large compression set and needs replacing within two years." (Xingyuan) The hub's gasket is silicone sponge, and it is at 16.7%. The PU foam gasket is at 15%, and it is leaking. The material is the difference between a joint that is one cycle from the floor and a joint that has crossed it.
Step 4, the fasteners and the corners. The lid's perimeter is 2 × (100 + 60) = 320 mm. Eight screws around 320 mm is 40 mm spacing, exactly at ATI Propel's limit of no greater than 40 mm. The spacing is right, and it is right by exactly nothing: seven screws would be 45.7 mm, and the lid would bulge between them, and the joint would leak in the middle of the span. And the corners: below 1.5× cord diameter the gasket wrinkles at a 90° bend, and a wrinkle always leaks. (Xingyuan) The hub's cord is 3.0 mm, so the groove corners need a radius of at least 1.5 × 3.0 = 4.5 mm. If the corners are drawn tighter, if the groove turns a sharp 90°, the gasket wrinkles at all four corners, and the wrinkles are four leaks, and the fastener spacing is irrelevant, because the corners leak first.
Step 5, the pressure. The hub is sealed at 20 °C, 101.3 kPa. The hub's spec is −20 °C to +60 °C, and the self-heating is 5 W in the sealed box, about a 25 °C rise, so at 60 °C ambient the inside is 85 °C. (Articles #7 and #50) At 85 °C, the pressure inside is 101.3 × 358.15 / 293.15 = 123.8 kPa, an overpressure of 22.5 kPa. At −20 °C, it is 101.3 × 253.15 / 293.15 = 87.5 kPa, a vacuum of 13.8 kPa. The full swing is 36.3 kPa, every cycle. The lid's area is 100 × 60 mm = 0.006 m². The overpressure pushes the lid with 22,500 × 0.006 = 134.7 N, about 13.7 kg of force, every hot day. The vacuum pulls it in with 13,800 × 0.006 = 83.0 N, about 8.5 kg, and the reversed differential at night pulls the gasket lip inward, and draws water through the weakest joint. (Xingyuan) The IP65 jet test is 30 kPa. The hot day is 22.5 kPa. The seal sees as much pressure from a hot day as from the jet test, and the jet test lasts three minutes, and the hot day lasts all afternoon, and there are two hundred hot days in two years.
Step 6, the fix, and the cost of finding it late. The fix is five numbers, and they are all in this article. The material: closed-cell silicone sponge, not PU foam, PU foam needs replacing within two years, and the hub's gasket is two years old. The compression: 30%, in a 2.1 mm groove, with a hard stop that prevents over-compression, "the seam geometry needs a controlled compression channel and a hard stop that prevents over-compression. A gasket that is crushed flat loses its memory and will fail after the first temperature cycle." (ATI Propel) The fill: 75%, in a 2.1 × 4.5 mm groove. The corners: at least 4.5 mm radius, moulded 45° corner pieces, splices away from the latches and the bottom centre. The vent: a Gore ePTFE membrane on the bottom of the base, under the lid's overhang, so the 36.3 kPa swing never loads the seal. And the fasteners: eight screws at 40 mm, torqued the same way every time, or a ninth screw, so the spacing is 35.6 mm and the margin is real.
The cost of finding it late is Article #40's cost escalation, 1× design, 5× prototype, 20× pilot, 50× production, 100×+ field, and Article #40's field failure: 4 of 10 hubs failed on days 19 to 23, 40% of the fleet. And Article #7's figure: 32% of outdoor edge-node failures are tied to faulty sealing, a vendor-guide industry-consensus figure, labelled as such. The seal is a set of numbers, and the numbers are cheap at the drawing stage, and they are 100 times more expensive in the field.
The decision test
Five questions. If the answer to any one of them is no, the seal is a material, not a set of numbers, and the decision is to design it as numbers, and to check them after two years of temperature cycling.
- What is the working compression, and what is it after two years? Not the design compression, the working compression, after the set. If the joint was designed at 30% and the gasket took a 40% set, the working compression is 20.5%. If the material is PU foam, it is 15%. The design number is the bench. The working number is the field.
- What is the groove depth, and what is its tolerance? The groove depth is what sets the compression, and a groove 0.1 mm too deep removes roughly a third of the compression on a 3 mm cord. If the drawing does not say the groove depth, or the groove depth was never measured, the seal was never designed.
- What is the fastener spacing, and what is the corner radius? The spacing must be no greater than 40 mm, and the corners must be at least 1.5× the cord diameter. If the spacing is 45.7 mm, the lid bulges. If the corners are sharp, the gasket wrinkles. A wrinkle always leaks.
- Does the box breathe, and where is the vent? The sealed box rises 20 K and carries about 6.7 kPa, and the hub's full swing is 36.3 kPa. If there is no vent, the seal holds the pressure, and the pressure is 134.7 N on the lid, every hot day. The vent belongs on the bottom, under an overhang, not on the flat top.
- Which joint is the worst, and is it the lid? The rating belongs to the whole enclosure, and the lowest-performing joint sets the result. The lid gasket is the one the buyer named. The cable gland is the one that is probably leaking. If the worst joint was never checked, the rating is a guess.
Objections
"The enclosure passed IP65, the seal is fine." The enclosure passed IP65 on the bench, at room temperature, on day one, for three minutes, at 30 kPa. The field is not the bench: the field is two years of temperature cycling, a 36.3 kPa pressure swing, a gasket that took a 40% set, and a groove that is 0.1 mm too deep. The bench test is a snapshot. The failure is a drift. The certificate says the seal was fine. The field says the seal is now working at 16.7%, and 15% is the floor.
"We used silicone, silicone doesn't take a set." Silicone takes less of a set than PU foam, and it lasts 10 to 15 years instead of two, but it takes a set, and the set is measured by ISO 815-1, and the set depends on the time and temperature of compression. The hub's silicone gasket took a 40% set in two years, and the working compression fell from 30% to 20.5%. The material is not the fix. The material is one of the numbers, and the numbers are the compression, the groove, the fill, the fasteners, and the vent.
"We'll just tighten the screws." Tightening the screws squeezes the gasket more, and squeezing it more takes a set faster, over-compression above 50% exceeds the material's yield limit, leading to permanent set and loss of long-term rebound capability. (Hinges Manufacturers) The screws are not a fix for a gasket that has taken a set. The screws are the reason the gasket took a set, if they were over-torqued. The fix is a new gasket, a checked groove, and a hard stop that prevents over-compression, not a tighter screw.
"The vent is a hole, it will let water in." The vent is a membrane, not a hole. A hydrophobic vent has a microporous structure with pores too small for water molecules to pass through, water has high surface tension, so it is more likely to sit atop the membrane than to permeate its pores. (Gore) The vent's own rating is IP65 for water jets, IP67 for 1 metre submersion for 30 minutes, and IP69K for high-pressure spray. (Gore) The vent does not let water in. The vent lets the pressure out, and the pressure is what pushes the water past the seal.
"Ultrasonic welding would solve it." Ultrasonic welding fuses the two halves into a single part and is the most reliable seal for sealed-for-life products. (ATI Propel) The hub is not sealed for life. It opens 25 times over a 10-year life, for battery service and calibration, and it is recalibrated, not replaced. The weld would solve the leak by making the product unserviceable, and the unserviceable product is a different product, and it is not the one the buyer sells.
Sources
- Sanjun, "IP65 Outdoor Enclosure: Sealing, Drainage & Gaskets"
What we used
"IP65 is not a property of a panel or a gasket. It is a property of a finished assembly, and it is set by the worst joint on that assembly." "IP65 covers dust-tightness and water jets only. It does not cover immersion, washdown or corrosion." "The rating belongs to the whole enclosure, so the lowest-performing joint sets the result." "Door gaskets need a working compression of roughly 25 to 35 per cent, set by groove depth rather than by feel." "A door seal is only as good as the groove it sits in. Groove depth, groove width and corner radius together determine how much the seal is squeezed when the door shuts, and that number is what makes or breaks the joint. For closed-cell sponge we work to roughly 25 to 35 per cent compression of free height. Compress it more and the gasket takes a permanent set, after which it stops pushing back and the joint opens; compress it less and the first thermal cycle is enough to lift it off." "Butt joints in a gasket run are planned leaks. They are sometimes unavoidable, and when they are, they belong at the top of the opening, sloped so water runs out rather than in, with the cut ends sealed. A butt joint at the bottom of a door is a gutter pointed at your electronics." "Fastener spacing on removable covers. Too few screws means uneven compression, and uneven compression is a gap." "The fifth is the one most often missed, and it is also the cheapest to get right at drawing stage. Adding screws later means a new tool, a new torque instruction and a new assembly step for a problem that was visible on a CAD screen."
- ATI Propel, "Waterproof Product Design Guide: IP Ratings and Seals"
What we used
"O-rings sit in a machined or molded groove and seal through controlled compression, typically 15 to 30 percent." "Screw-compressed seams use a continuous gasket pressed by evenly spaced fasteners; the rule of thumb is fastener spacing no greater than 40 millimeters to prevent gasket bulging between screws." "Snap-fit seams with integrated seals are common for low-cost consumer products but rarely exceed IPX4 reliably. Ultrasonic welding fuses the two halves into a single part and is the most reliable seal for sealed-for-life products like wearables and medical disposables." "A sealed enclosure is not a static system. When the product warms up in sunlight or cools down in a refrigerator, the air inside expands and contracts. The pressure difference can be large enough to push water past a seal, distort a membrane, or pop a battery door. This is why almost every serious waterproof product includes a small breathable vent, typically a Gore-Tex style ePTFE membrane, that lets air pass but blocks liquid water." "IPX4 means splashes from any direction. IPX7 means temporary immersion up to one meter for 30 minutes." "A vent on the bottom of a wearable, recessed under a small overhang, is far more reliable than a vent on a flat top surface." "Whichever approach you choose, the seam geometry needs a controlled compression channel and a hard stop that prevents over-compression. A gasket that is crushed flat loses its memory and will fail after the first temperature cycle."
- Hinges Manufacturers, "Compression Latches for IP65/IP66 Enclosures: Engineering Guide"
What we used
"Compression Latches differ fundamentally from standard Cam Latches by their ability to provide a 3mm–6mm axial stroke after closing. This stroke generates a controlled Preload, forcing the gasket into its effective compression range (20%–40%), thereby ensuring compliance with IP65/66 and NEMA 4/4X standards." "Under-compression (<15%): Fails to fill microscopic irregularities in the door panel, creating capillary siphon channels." "Over-compression (>50%): Exceeds the material's yield limit, leading to Permanent Set and loss of long-term rebound capability." "Reference Standard: ASTM D395 or ISO 815 (Standard Test Methods for Rubber Property, Compression Set)." "Compression Stroke: Recommended value ≥ 3mm." "With high-strength material composition (5,000N compression / 3,000N tensile / HRC 50–55 hardness), compliance with IP65/66 and NEMA 4/4X standards, and a 100,000-cycle rated service life."
- ISO 815-1:2019, "Rubber, vulcanized or thermoplastic, Determination of compression set, Part 1: At ambient or elevated temperatures"
What we used
"This document specifies methods for the determination of the compression set characteristics of vulcanized and thermoplastic rubbers at ambient (one method) or elevated temperatures (three methods, A, B, and C, depending on the way the test piece is released at the end of the test)." "The methods are intended to measure the ability of rubbers of hardness within the range 10 IRHD to 95 IRHD to retain their elastic properties at specified temperatures after prolonged compression at constant strain (normally 25 %) under one of the alternative sets of conditions described." "When rubber is held under compression, physical or chemical changes that prevent the rubber returning to its original dimensions after release of the deforming force can occur. The result is a set, the magnitude of which depends on the time and temperature of compression as well as on the time, temperature, and conditions of recovery."
- Elastotech, "Rubber compression set"
What we used
"Compression set (%) = [(t₀ − tᵣ) ÷ (t₀ − tₛ)] × 100, t₀: original specimen thickness; tᵣ: recovered thickness after the specified recovery period; tₛ: spacer thickness used to hold the compressed specimen." "Suppose a hypothetical specimen starts at 10 mm, is held at 7.5 mm and recovers to 9.5 mm. The original compression is 2.5 mm and the unrecovered thickness is 0.5 mm. The result is (0.5 ÷ 2.5) × 100 = 20%." "ASTM D395 includes compression-set test methods using constant force or constant deflection. ISO 815-1 addresses compression set at ambient or elevated temperatures. A standard number alone is not enough to compare two percentages fairly."
- Engineers Edge, "Groove Design for Metric ISO Size O-Rings per. ISO 3601-2"
What we used
"An initial compression (squeeze) of the O-Ring in the groove is essential to ensure its function as a primary or secondary sealing element." "Dynamic applications: 6 to 20%. Static applications: 15 to 30%."
- Projects House, "O-Ring Selection and Gland Design: Squeeze, Groove Sizing, and Materials"
What we used
"Squeeze. Aim for roughly 15 to 30 percent compression of the ring cross-section. Static seals tolerate the upper end; dynamic seals live near the bottom." "Gland fill. The ring should occupy about 75 to 90 percent of the groove volume, leaving room for thermal expansion and fluid swell. A completely filled groove hydraulically locks and can split the housing." "Stretch. Installed stretch on an inner diameter should stay within a few percent." "The chain from lid to base to boss to fastener determines the squeeze that actually occurs, so compute the worst case, not just the nominal, the method is set out in tolerance stack-up analysis."
- Xingyuan, "Wall-Mount Enclosures to IP66: Gasket, Knockouts and Wall Load"
What we used
"a sealed 600×400×250 box rising 20 K carries about 6.7 kPa (ΔP = P·ΔT/T), and the reversed differential at night pulls the gasket lip inward." "Below about 15% ordinary flatness error opens the joint; above 45% the gasket is crushed, loses rebound and takes a permanent set in a few thermal cycles. Outdoors the window is 25–40%." "Closed-cell silicone sponge, 10–25 Shore A: the default, best rebound, UV and ozone stable." "PU foam: convenient for irregular shapes, but takes a large compression set and needs replacing within two years." "Width sets fill ratio, gasket section over groove section, at 70–85%. Below 70% the gasket walks; above 85% it extrudes and lifts the panel. An 8 mm cord in a 5.5 mm deep groove gives 31% compression, which is right; at 10 mm width the fill ratio is 91%, at 12 mm it is 76%." "Corners: below 1.5× cord diameter the gasket wrinkles at a 90° bend, and a wrinkle always leaks." "Coating eats compression: 60–80 µm of powder inside the groove removes 0.06–0.08 mm of depth." "The last path is rarely budgeted: a sealed box is a breathing vessel. The better the seal, the larger the swing, and the harder it pulls water through its weakest joint."
- xkyhardware, "IP65 vs IP67 Enclosure Seal Design"
What we used
"A seal is a set of numbers, not a material: 20–30% compression, 70–90% groove fill, 1.6 µm Ra seats, fasteners every 50–100 mm." "Compression: target 20–30% for solid cord. Below 20% the gasket never reaches its sealing stress; above 30% it takes a permanent set and leaks within months." "Fill ratio: keep the groove cross-section 70–90% filled by the gasket." "Seat finish: machine gasket seats to 1.6 µm Ra or better, and check flatness after machining, a seat that is smooth but dished will still leak." "Groove tolerance: depth is what sets compression. A groove 0.1 mm too deep removes roughly a third of the compression on a 3 mm cord." "EPDM: roughly −40 °C to 120 °C." "Space fasteners every 50–100 mm on a machined aluminium flange, closer on thin or flexible lids."
- JS Precision, "Custom Sheet Metal Enclosure DFM: IP65/IP66/IP67 Sealing"
What we used
"Custom sheet metal enclosure IP sealing depends on three variables: 25%–35% gasket compression, rigid flange geometry, and fastener pitch tight enough to prevent panel bowing." "IP65 resists low-pressure water jets, IP66 resists powerful jets, and IP67 survives temporary immersion in 1 m of water for 30 minutes." "Achieving IP65 prevents dust entry and resists 12.5 L/min water jets at 30 kPa, whereas IP66 withstands 100 L/min heavy seas at 100 kPa." "Fastener pitch ≤ 75 mm for IP65 (t ≥ 1.5 mm), ≤ 50 mm for IP66 (t ≥ 2.0 mm), ≤ 120 mm for IP54." "The key to achieving a high level of protection is to use double folded rigid flanges to counteract the bending of the plate under the fastening preload force, ensuring that the uniformity of compression throughout the entire circumference is within 10%."
- Gore, "The Impact of Temperature Cycling on Sealed Enclosures" (case study)
What we used
"Both enclosures are exposed to a temperature increase from 20°C (68°F) to 50°C (122°F) in a time frame of 90 seconds (control volume = 4000 cm3). When heat is applied, the resulting increase in internal pressure within enclosure B causes the balloon to inflate. In contrast, enclosure A maintains stable internal conditions because the GORE® Protective Vent enables air exchange, allowing pressure to equalize with the external environment." "Exposed to a 90-second, rapid cooling cycle from an ambient 20°C (68°F) to -5°C (23°F)... The decreased air pressure in enclosure B creates a significant vacuum, causing the balloon to deflate." "In a real-world situation, contaminants such as liquid, dirt and particles can be drawn into a device, causing failure to internal components."
- Gore, "GORE Protective Vents Adhesive Series VE8, VE7 and VE9" (datasheet)
What we used
"Outdoor enclosures are continuously exposed to harsh environments such as rainstorms, dust, sand and high winds. During changing environmental conditions, pressure can build inside a sealed enclosure, putting stress on seals." "Vents equalize pressure and reduce condensation by allowing air to flow freely into and out of sealed enclosures." "With proven performance for more than 25 years... over 200 million installations worldwide." "Typical airflow (dp = 70 mbar): VE8 3300 ml/min/cm², VE7 290 ml/min/cm², VE9 1150 ml/min/cm²." "A hydrophobic vent from Gore has a microporous structure with pores too small for water molecules to pass through. Water has high surface tension, so it's more likely to sit atop a vent membrane than to permeate (or 'wet-out') the membrane's pores."
- Demo arithmetic verified in Python before writing: design compression (3.0 − 2.1) / 3.0 = 30%, squeeze 0.9 mm; groove depth for 25–35% on a 3.0 mm cord = 1.95–2.25 mm; cord area π × 1.5² = 7.07 mm², groove area 2.1 × 4.5 = 9.45 mm², fill = 74.8%. Compression set: TEMCO's numbers 10.0 → 7.5 → 9.0 mm = (10.0 − 9.0) / (10.0 − 7.5) = 40%; the hub's gasket compressed 0.9 mm, a 40% set loses 0.36 mm, free height 2.64 mm, working compression (2.64 − 2.1) / 2.64 = 20.5%. Groove 0.1 mm too deep (2.2 mm): (2.64 − 2.2) / 2.64 = 16.7%. For 15% working compression with a 2.1 mm groove: free height 2.1 / 0.85 = 2.47 mm, lost 0.53 mm, set 0.53 / 0.9 = 58.8% ≈ 59%. Lid perimeter 2 × (100 + 60) = 320 mm; 8 screws → 40 mm spacing; 7 screws → 45.7 mm. Corner radius minimum 1.5 × 3.0 = 4.5 mm. Pressure: sealed at 20 °C (293.15 K), 101.3 kPa; at 85 °C (358.15 K) → 123.8 kPa, +22.5 kPa; at −20 °C (253.15 K) → 87.5 kPa, −13.8 kPa; full swing 36.3 kPa; 20 K rise from 300 K → 101.3 × 20 / 300 = 6.75 kPa (Xingyuan's "about 6.7 kPa"
What we used
). Lid area 0.006 m²: overpressure force 22,500 × 0.006 = 134.7 N (≈13.7 kg); vacuum force 13,800 × 0.006 = 83.0 N (≈8.5 kg). Hub volume 240 cm³. Xingyuan's 8 mm cord in a 5.5 mm groove = 31.25% compression; fill 91% at 10 mm width, 76% at 12 mm