Two tests, two different questions
Every ball valve leaves a works with a test record, and the record nearly always names a standard, a pressure and a result. What it frequently leaves out is the medium, and the medium is what decides which failures the test was capable of finding. Water and air are not interchangeable versions of the same check. Water is almost incompressible, its viscosity and surface tension let it bridge a passage that no gas would pass, and it does not dissolve into a polymer seat. Air is compressible, it stores energy, it forms no meniscus worth the name, and it passes through materials that water cannot wet. The same valve, at the same pressure, tested with the two media, can produce two different answers and both can be honest.
That is the whole argument in one paragraph, and it is worth stating plainly because most test disputes are not about whether a valve leaked. They are about what a piece of paper proved.
What the hydrostatic seat test can prove
API 598 applies a shell test at 1.5 times the rated pressure and a seat test at 1.1 times, and the two answer different questions. A water seat test that passes proves that the valve holds liquid at that pressure to the acceptance class, with the seat loaded in the direction and to the magnitude the design intends, and that nothing in the body, the body joint or the closure welds failed under 1.5 times. That is a real proof and a necessary one, and it is a proof about liquid. The Rate A and Rate B comparison on this site deals with the acceptance classes themselves, which class a soft seated valve should be held to and how the shell and seat tests differ in what they load; this article is about the fluid, not the class.
What a water test does not prove is that the valve will isolate gas. Nothing about passing it says anything about a gas molecule crossing the same band, and treating a hydrostatic pass as a gas tightness proof is the single most common reading error in valve acceptance.
Three reasons a valve passes on water and leaks on gas
The first is liquid bridging. A scratch, a pore in a coating, a score across the band: water forms a meniscus across a passage of a certain size and needs a pressure difference to be pushed through it, while gas does not form one. A path that is a leak to a gas can be, for a while, a seal to a liquid, and it stays a seal for exactly as long as the meniscus holds.
The second is permeation. Gas dissolves in and diffuses through polymers in a way water does not, so a soft seat that is bubble tight to water can show a gas reading that comes through the seat material itself rather than past it. That is not a mechanical leak path, and it is why a gas result on a PTFE seated valve has to be read together with the seat material and the hold time rather than as a verdict on the sealing geometry. It also explains a conversation that repeats in every gas project: the valve is not leaking, it is permeating, and the difference decides whether the answer is a different seat or a different acceptance criterion.
The third is the seat's own deformation. A soft seat seals by conforming. Under a water test at ambient it conforms to whatever the band gives it and the test passes. In gas service at temperature the same seat has a different modulus and a different creep rate, and a contact that was conforming becomes a contact with a gap that a gas molecule crosses and a water meniscus does not.
Why a pneumatic test is a different procedure, not an extra line
A compressed gas volume holds stored energy, so a failure under a gas test is a release of that energy rather than a drip. That single fact explains the shape of every gas test procedure in the industry. The test pressure is restricted below what the same valve would take in water, the gauges are read from a distance, the area is barricaded, pressurisation is staged rather than applied in one step, and nobody stands in front of a flange or a body joint while the pressure is on.
It also explains why a factory defaults to water: a water test is safer, cheaper and faster, and a gas test is a separate operation with its own risk assessment and its own record. So when a specification asks for a pneumatic seat test, it is not asking for the same test with a different fluid. It is asking for a different pressure basis, a different acceptance reading and a different safety envelope, and a quotation that does not show it understood the difference is quoting the words rather than the work.
Reading a bubble
Bubble testing is done two ways, by immersion or with a film of leak detection liquid painted on the joint. Immersion is the more sensitive of the two and it puts water on the valve, which matters later. The bubble itself carries information that is usually discarded. Where it appears identifies which seal is passing: a bubble at the body joint is a gasket, at the stem it is packing, at a plug it is a threaded connection, at the seat it is the band, and the difference decides whether the valve goes back to assembly or to the seat bench.
How the rate behaves over time says what kind of path it is. A burst that decays over the first minute is trapped air leaving the cavity, and a rate that stays constant is a leak. And a bubble is a location, not a quantity, so a certificate reading "no bubbles observed" without the observation time, the medium and the side under test has recorded nothing that can be compared with another valve or another batch.
Pressure decay, and the reading temperature can fake
A decay test closes a volume, pressurises it, and watches the gauge. Charging the volume heats the gas by compression, the gas then cools towards ambient, and the pressure falls with no leak at all. A decay test without a stabilisation period is measuring the thermal history of the test rather than the tightness of the valve, and it will fail good valves in the morning and pass leaking ones in the afternoon, depending on the weather in the shop.
Two more effects belong with it. Gauge resolution against the volume, because the bigger the trapped volume the smaller the pressure change that a given leak produces, so a valve tested with a length of pipe attached is easier to pass and much harder to judge. And the direction of the leak, because a decay test on the whole valve cannot tell a seat leak from a body joint leak, which is the one thing a bubble test does well. The remedy for both is procedural: a reference volume, a stabilisation period, a recorded ambient temperature, and a bubble test when the answer has to include a location.
Nitrogen and helium: what each one buys
Nitrogen is dry, inert, cheap and available on any site. It makes a whole valve gas test practical, it leaves no water behind, and it finds gross leaks. It also hides small ones, because a nitrogen molecule is not a small molecule and a decay reading taken on nitrogen is a coarse instrument. Helium is the tracer of choice when the question is a total leak rate rather than a visible bubble: it is light, it diffuses quickly, and a mass spectrometer reads it at a level a bubble test cannot approach.
Two cautions travel with it. Helium permeates some elastomers and some glasses, so a rising background is not automatically a leak through the seal, and a test that does not establish a background first will report the instrument rather than the valve. And a tracer test only answers for the side the tracer is on, so each seat has to be tested in turn, exactly as a water seat test is run on each side.
Four media, four different findings
Laid side by side, the four things that can be put inside a valve during a test are not ranked from worse to better. Each one answers a question the others cannot, and each one leaves something behind that the next operation has to deal with.
The medium you leave behind
A hydro test leaves water, and water in a ball valve sits in the body cavity, in the seat pockets and around the stem, which are the three places with the least airflow and the last to dry. In stainless valves the chloride in test water is a pitting risk that a rinse does not always remove, which is why test water chemistry is controlled for austenitic and duplex bodies. In carbon steel the residue becomes rust in the cavity, and it arrives at site as debris sitting on the seat. In cryogenic service it becomes ice in a seat pocket, and ice in a seat pocket is a leak. In oxygen and dry chlorine service water is not a residue but a hazard, and the valve has to be dried and capped to the same procedure that governs its cleanliness.
That is why "tested, drained and dried" is a line in a specification rather than an assumption, and why the caps that go on the flanges after test are part of the delivery. A gas test avoids the water problem and creates a different one: the valve is left at pressure, and the order in which it is depressurised, and from which side, is a step in the procedure rather than an afterthought.
Why a line test is not a valve test
When a pipeline section is hydrotested, the block valve at the end of the section is usually part of the test. It sees the full test pressure on one face with atmosphere on the other, which is a load case it may never see in service. It may be used as the blind that closes the section, which loads the seat in a direction the design never intended. And it may be left partly open while the line is flushed, which sends mill scale and weld slag through the seat at speed. Then the same valve is signed off as tested, on the strength of a test that was measuring the pipe.
Three traps follow. A test pressure above the valve's cold working pressure at the test temperature, because the rating is a pressure and temperature table and not a pressure table. Debris from the line settling on the band during the test and being pressed into it by the seat, which shows up weeks later as a leak that cannot be reproduced on the bench. And a seat loaded from the back, which can extrude a soft seat in a direction the design never intended. A valve that passed a line test and leaks at first operation has usually been tested rather than proven.
What to write into the purchase order or the ITP
Seven lines, and the first is the one that gets omitted. The test medium for each test, because the shell and the seat test do not have to use the same one. The pressure and its basis, taken as 1.5 times the rated pressure for the shell and 1.1 times for the seat to API 598, calculated from the cold working pressure at the test temperature. The duration and the acceptance class, with ISO 5208 named where the class is what the buyer is actually buying. Which side of each seat is tested, and in what order. Whether a gas test is required in addition, and if so the protection that goes with it. Drying and end protection after the test, with caps and desiccant stated. And the record itself: medium, pressure, duration, class, result and witness.
J-VALVES tests every valve at 1.5 times rated pressure on the shell and 1.1 times on the seat to API 598 and API 6D, for ball valves from 1/2 to 56 inch (DN15 to DN1400) in ASME Class 150 to 2500, with a written pressure test procedure and acceptance criteria issued with the order. Gas seat tests are run to the procedure agreed in the ITP, and the underlying standard requirements are set out in our summary of API 6D and ASME B16.34, with EN 10204 3.1 certificates available for body, ball and stem.
Frequently Asked Questions
Can a ball valve pass the hydrostatic seat test and still leak gas in service? Yes. Liquid bridging, permeation through the seat polymer and the difference in seat deformation between a water test at ambient and gas service at temperature all allow a valve to hold water at 1.1 times rated pressure and pass gas in the line. A hydrostatic test proves liquid tightness and structural integrity; it is not a gas tightness proof.
Is a pneumatic test more accurate than a hydrostatic test? Not more accurate, different. It finds gas-path leaks that a water meniscus can bridge, and it cannot prove structural strength the way a shell test at 1.5 times does. Because a compressed gas stores energy, it is run at a restricted pressure with barricades, remote gauging and staged pressurisation, which is why it is a separate operation rather than an option on the same test.
Bubble test or pressure decay, which should I ask for? The bubble test locates the leak, at the seat, the body joint, the stem or a plug, and a pressure decay test gives a single number for the whole trapped volume without saying where it went. For a valve the bubble test is usually the more useful factory record and the decay test is the field instrument.
Should a stainless ball valve be hydrotested with the water available on site? Only with the chloride controlled. Untreated site water can leave chloride in the body cavity and the seat pockets, which is a pitting risk that a rinse does not reliably remove, and the valve should be drained, dried and capped after the test. On a cryogenic valve the same residue becomes ice, and on a metal seated valve tested for gas service it becomes a leak path the seat cannot close, which is the clearance problem described for metal seated construction.
