Valve Seat Leakage Classes: ISO 5208 Rates And What Metal Seats Can Deliver

More procurement disputes in the valve trade start with one line on a data sheet than with any price negotiation: the seat leakage requirement. A buyer writes "bubble tight" or "zero leakage" or "Class IV shut-off", the manufacturer reads it as something else, and six weeks later a valve that both parties believe complies fails the acceptance test in front of a witness. The problem is almost never the valve. It is that "tight" was never defined in units, fluid and test conditions.

This article puts numbers on the leakage classes that actually govern industrial valve procurement. It explains the ISO 5208 Rate A to G ladder with a full conversion table, shows how API 598, API 6D, MSS SP-61 and ANSI/FCI 70-2 map onto it, and then answers the question buyers really care about: what can a metal-to-metal seat honestly achieve, and what has to be engineered into the valve to get there. The testing procedures themselves are covered in our API 598 pressure testing guide; here the subject is the acceptance numbers and the sealing engineering behind them.

1. Why Leakage Class Is a Contract Number, Not a Valve Property

A leakage class is not a physical property of a valve like body wall thickness. It is an agreement: a maximum measured leakage, with a specified fluid, at a specified pressure, over a specified duration, measured in a specified way. Change any one of those five parameters and the same valve can pass one class and fail another without anything about the valve changing.

Three consequences follow, and they explain most specification mistakes. First, a leakage rate quoted without its test conditions is meaningless; "Rate A" proven at 6 bar of nitrogen over 60 seconds says nothing about hold-tight performance at operating pressure over a shift. Second, tighter is not automatically better: each step down the ladder costs real money in lapping time, hardfacing, actuator torque and test hours, and a class the process does not need is budget spent for nothing. Third, the class you specify must be one the product standard actually accepts, or the acceptance criteria conflict and the inspector arbitrates by choosing whichever is stricter, usually at your expediting schedule's expense.

2. One Table That Matters: the ISO 5208 Rates A to G

ISO 5208, "Industrial valves - Pressure testing of metallic valves", is the reference behind most modern leakage specifications, and it is the standard that API 6D, EN 12266-1 and most European product standards point to. Its Table 4 defines ten closure-test leakage rates. The elegant part is the scaling: from Rate AA upward, the maximum allowable leakage is expressed as a factor multiplied by the nominal size DN, so one table covers a DN 15 needle-seated globe valve and a DN 1200 pipeline ball valve with the same arithmetic.

Three notes from the standard change how the table is used in practice. The rates apply only when the test fluid discharges to atmosphere, not to a closed volume. The rate that applies is either the one named in the valve product standard or a stricter one the purchaser writes into the order - the standard gives the buyer the right to tighten, never to loosen. And Rates A through G correspond one-to-one with EN 12266-1, so a European project specification and an ISO-based one can be reconciled by letter without conversion.

A worked example fixes the scale in mind. On a DN 200 valve, Rate D liquid allows 0.1 × 200 = 20 mm³/s, which is 1.2 mL/min, about 19 drops of water per minute at the customary 16 drops per mL. Rate C on the same valve allows 6 mm³/s, roughly 6 drops per minute. Rate AA gas allows 36 mm³/s of air. Those are the numbers an inspector compares against the drip counter or the bubble tube.

3. Liquid Allowances versus Gas Allowances: the Hundredfold Gap

The single most surprising column in the ISO table is the gas one, because the gas allowance is not a little looser than the liquid allowance - it is two to three orders of magnitude looser. At Rate D, a DN 200 valve may pass a water test allowing 1.2 mL/min while the same rate on a gas test tolerates 30 × 200 = 6000 mm³/s, which is six litres per minute. Even Rate CC gas, an intermediate class, permits 22.3 × 200 = 4460 mm³/s on the same valve.

The reason is measurement physics, not leniency. Water leaking past a metal seat builds visible drops at flow rates thousands of times smaller than the flow rates at which air becomes countable as bubbles, because surface tension, wetting and the viscosity ratio all work in the observer's favour on the liquid side. A gas test therefore detects what a water test cannot, which is why every serious tightness requirement for hazardous, flammable or valuable gas service is written as a gas rate - and why a supplier who tests a metal-seated valve with water only and hands it over against a gas-tight requirement has not actually tested the property you bought.

The practical rule for buyers: when the process fluid is a gas or a flashing liquid, specify the gas rate by letter and ask for the gas seat test to be witnessed. When the process is a clean liquid, the liquid rate is the governing number and the water test is representative. When the process is a dirty liquid or a slurry, remember that neither number predicts much, because the class was defined on clean test conditions - the valve you get on day one is the best it will ever be.

4. How API 598, API 6D, MSS SP-61 and FCI 70-2 Map onto ISO 5208

Four other documents appear in purchase orders, and each uses its own units. Mapping them onto the ISO ladder settles most acceptance disputes before they start.

Two traps hide in this table. The first is MSS SP-61: a supplier offering it where the specification calls for API 598 is offering a measurably looser test, and it is a legitimate-looking substitution that passes unnoticed in a stack of certificates. The second is Class IV appearing on an isolation valve data sheet. FCI 70-2 was written for throttling control valves, and its capacity-percentage classes are defined against rated Cv - a Class IV valve with a Cv of 100 is allowed 0.01% of that, about 0.01 gpm or 38 mL/min, which is looser than ISO 5208 Rate D liquid on the same valve. "Class IV shut-off" on a pipeline ball valve is therefore not a tightness claim at all; it is a category error that belongs to a different standard. Where the intention is a tight metal-seated isolation valve, specify the ISO rate.

5. Rate A Means "No Visible Leakage" - and That Depends on Duration and Eyesight

Rate A looks like the strictest class, and in most commercial contexts it is. But it is defined visually: no leakage detectable by eye - no weeping, no drops, no bubbles - for the duration of the test. That definition has edges worth knowing before a witness signs a test report.

The first edge is duration. ISO 5208 closure test minimum durations are 15 seconds up to DN 50, 60 seconds from DN 65 to 150, and 120 seconds at DN 200 and above. API 6D's seat test durations are longer - 2, 5 and 10 minutes across the same size bands - so a Rate A result under API 6D is a stronger statement than the same letter under a bare ISO 5208 test. The second edge is the pressure: the gas closure test runs at 6 bar ± 1 bar, or 1.1 times the cold working pressure if that is higher, and Rate A at 6 bar is a much weaker claim than demonstrated tightness at operating pressure. The third edge is the human one: "visually detectable" assumes clean water, good light and a trained eye, which is why reputable shops soak the downstream end in a bubble bath or use a paper-towel wipe check on the liquid side rather than trusting a glance.

None of this makes Rate A a weak class - on soft-seated valves it is routinely achieved and routinely witnessed. It means that for the services where invisible leakage still matters, the specification should reach one step further into the ladder, to Rate AA, which replaces the observer with a number.

6. Why Metal-to-Metal Seats Leak at All

A soft seat seals because an elastomer or PTFE yields at low contact stress and conforms to every machining mark on the counterface; the material fills the leak path. A metal seat has no such luxury. Metal-to-metal contact happens at asperities - the peaks of the surface texture - and between those peaks are channels that a fluid at pressure will find. The seal is only as good as the fraction of the nominal seat area that is actually in contact, and that fraction is bought with surface finish, flatness, contact stress and geometry.

The numbers involved explain why the classes on metal seats cluster where they do. Machined seats leave finish around Ra 1.6 to 3.2 microns, which supports Rate D-class liquid tightness at best. Lapping brings the finish to Ra 0.2 to 0.4 microns and, more importantly, improves flatness so that the contact band closes circumferentially; that is the difference between Rate C and Rate D behaviour. Grinding marks oriented radially or spirally act as prefabricated leak paths, so a properly lapped seat shows concentric lay with no visible spiral. And because neither surface can yield, the pair needs real unit contact stress to flatten asperities elastically - which is why metal-seated valves need more closing force than their soft-seated equivalents, and why an actuator sized without torque margin converts a passing test bench valve into a leaking installed one.

This is also the boundary between the two seat philosophies. A soft seat buys Rate A with a polymer and loses it in fire, temperature or abrasion; a metal seat keeps its properties through fire and heat but can never match polymer conformance. The engineering question is never "which seat is tighter" in the abstract - it is which class the service needs, at which temperature, and what the seat must survive to still be tight next year. Our triple-offset butterfly valve article covers the geometry side of that trade-off in detail.

7. Six Ways Manufacturers Buy Tightness on a Metal Seat

When a specification demands more from a metal seat than standard machining delivers, the manufacturer reaches for one or more of the following levers, in roughly ascending order of cost.

1. Hardfacing the seat and disc faces. Stellite 6, nickel-boride or tungsten carbide overlays raise the surface hardness so that asperities flatten under load instead of galling, and they hold finish through handling and thermal cycling. Hardfacing alone does not make a seat tight - but every lapped tight seat on a severe-service valve sits on a hardface.

2. Lapping and bluing verification. Fine lapping compound on a matched lap, checked by bluing transfer: the acceptance target is a continuous circumferential contact band of about 80% or better with no spiral marks. This single operation moves a metal seat from Rate D territory toward Rate C and, with good geometry, a passing gas test at 6 bar.

3. Wedge and cam geometry. A seat pair that gains contact stress from mechanical interference rather than actuator force alone seals more predictably. The flexible wedge in a gate valve wedge design, and the third offset of a triple-offset butterfly, both exist for this reason: they convert a defined movement into a controlled, repeatable seating load.

4. Compliant metallic elements. Laminated seat rings with graphite interlayers, spring-energised seat bands and floated seat rings give the metal assembly a small amount of the conformity a soft seat gets for free. They recover from thermal movement and minor debris, which is why they are standard on high-temperature butterfly and trunnion ball valves.

5. Closing force margin. Tightness on a metal seat scales with unit contact stress, so seat leakage class is an actuator question as much as a seat question. The torque and thrust margins involved are quantified in our pneumatic actuator sizing guide; the short version is that a metal-seated valve tested with a hand lever and shipped with an under-torqued actuator will not repeat its factory result.

6. Backup sealing for fire and emergency. On soft-seated trunnion and ball valves, a secondary metal seat or graphite fire seal takes over after an API 607 fire event. The fire-test acceptance leakage after the burn is, by design, far looser than Rate A - the class to quote for post-fire performance is the fire test standard's own criterion, not an ISO rate.

8. What Each Valve Family Can Honestly Promise

Bringing the classes and the seat engineering together, the table below is the realistic answer to "how tight can it be" for common valve families at ambient test conditions. It is a screening guide, not a warranty - the governing document is always the manufacturer's test data for the actual size, class and trim.

One cross-reference belongs here. Buyers who see "zero leakage" quoted for hydrogen or other hazardous-gas duty should look at the whole document, not just the seat line: hydrogen valve specifications in practice pair an ISO seat rate with fugitive-emission requirements on the stem, per ISO 15848-1, and pipeline hydrogen valves under API 6D Annex M add tightness clauses of their own - SLH3 and SLH4 hydrogen service, for example, accepts no more than five times Rate C on metal seats. Seat leakage and stem leakage are different paths with different standards, and our ISO 15848-1 certification page covers the stem side.

9. Specification Language: What to Write, What to Reject

Everything above compresses into a handful of drafting rules for the seat leakage line of a purchase order.

• Write the rate, the fluid and the standard, not an adjective. "Seat leakage: ISO 5208 Rate D, liquid, per API 6D" is complete. "Bubble tight", "zero leakage" and "leak-free" are not - each has been read at least four ways in real disputes.

• For gas or flashing-liquid service, specify a gas rate. Rate A or AA gas with a witnessed nitrogen test is the honest version of "bubble tight".

• Name the product standard and let it set the default, then tighten deliberately. If API 6D governs, metal seats start at Rate D liquid; tightening to Rate C is a real engineering request with real cost and should be made consciously, per size and per valve.

• Do not import control-valve classes into isolation valves. Class IV and Class V belong to FCI 70-2 datasheets. If a design institute copies them onto an isolation valve, translate back to the ISO rate before ordering.

• Match the test to the claim at witness stage. A Rate A gas seat test witnessed at 6 bar with the duration recorded is worth more than any certificate language; ask that measured leakage be recorded even when the class is visual - the number costs nothing and becomes the baseline for in-service condition monitoring.

• Agree the fire-case expectation separately. After a fire event, quoting Rate A is meaningless; reference the fire test standard's acceptance instead, so the two criteria cannot be confused in an incident investigation.

The last rule is the one that pays repeatedly: record the actual measured leakage at FAT, class A or not. A valve that wept 3 drops per minute at Rate D at the factory has a fingerprint; two years later, 40 drops per minute is a trend you can see, and "the seat is wearing" becomes a measurement instead of an argument.

10. Closing the Loop between Class and Design

Leakage class is where valve design, manufacturing discipline and procurement language meet. The ISO 5208 ladder gives every party the same ruler: ten rates, one DN factor, one table. API 598 and API 6D point at that ruler from their own procedures, FCI 70-2 serves a different valve population on a different basis, and MSS SP-61 sits measurably looser than API 598 where the two look interchangeable. The buyer's job is to pick the letter, the fluid and the test conditions in one sentence; the manufacturer's job is to reach that letter with hardfacing, lapping, geometry and closing force rather than with adjectives. When both jobs are done, the acceptance test stops being a negotiation and becomes what it was always meant to be: a measurement.

What is the difference between ISO 5208 Rate A and Rate AA? 

Rate A is a visual criterion: no leakage detectable by eye - no weeping, drops or bubbles - for the duration of the test. Rate AA replaces the observer with a number: 0.006 times DN mm3/s on a water test and 0.18 times DN mm3/s on a gas test. On a DN 100 valve, Rate AA gas allows 18 mm3/s of air, measured rather than judged. Rate A is routinely achieved on soft-seated valves; for metal seats and for services where invisible leakage still matters, Rate AA is the honest specification because it is verifiable.

Is bubble tight the same as ISO 5208 Rate A? 

In practice yes, but as specification language it is incomplete. Bubble tight usually means no visible bubbles on a gas seat test, which matches Rate A only if the test pressure, duration and method match the product standard. ANSI/FCI 70-2 Class VI is also bubble-based but is not zero: an 8-inch port is allowed 45 bubbles per minute, and it is a control-valve class defined on a different basis. Write the rate, the fluid and the standard instead - for example: seat leakage ISO 5208 Rate A, gas, per API 6D.

Can a metal-to-metal seat really achieve Rate A? 

On the liquid side, frequently: a lapped hardfaced seat with a continuous circumferential contact band of 80% or better will show no drops in a water test. On the gas side, Rate A at 6 bar is achievable but it is a demonstrated result for that specific size and build, not a generic property of metal seats - conformity that a polymer gets for free has to be bought with hardfacing, lapping, geometry and closing force. Any Rate A gas claim on a metal-seated valve should be backed by a witnessed nitrogen test on the actual valve, and measured leakage recorded even though the class is visual.

What leakage class should a pipeline valve specification use? 

Let the product standard set the baseline and tighten deliberately. API 6D requires Rate A (no visible leakage) for soft-seated and lubricated plug valves, and for metal-seated block valves sets a Rate D liquid basis - the 24th edition text is Rate D with a note that special applications may require less, and the current edition wording cites Rate CD - while metal-seated check valves are held to Rate E (liquid). For hazardous or valuable gas service, specify a gas rate by letter, Rate A or Rate AA, with a witnessed nitrogen seat test, and record the measured leakage as a baseline.

Why did my valve pass the water seat test but leak on gas? 

The gas allowance in ISO 5208 is two to three orders of magnitude looser than the liquid allowance at the same rate letter, but gas is also a far more sensitive search fluid: air finds and flows through asperity channels that water bridges with surface tension. A metal seat can be genuinely tight to water at Rate D and still bubble on the 6 bar gas test. If the service fluid is a gas, a flashing liquid or anything where seepage counts, the seat test that predicts service behaviour is the gas test - specify the gas rate and test with nitrogen before the valve ships, not after it is installed.