Resilient-seated butterfly valves must show zero visible leakage in both the liquid and gas seat tests under API 598, at every NPS size, which is the standard working definition of bubble-tight [S2][S4].
Metal-seated butterfly valves are graded by size: NPS ≤ 2 is effectively bubble-tight, NPS 2 to 6 is capped at 12 liquid drops/min and 24 gas bubbles/min, NPS 8 to 12 at 20 drops/40 bubbles, and NPS > 12 at 28 drops/56 bubbles [S1][S2]. The closure test itself is run at 1.1× the rated pressure in water and at roughly 80 psig (≈5.5 bar) in air or nitrogen [S1][S4][S8].
What "bubble-tight" actually means under API 598
API 598 defines a "0 drop" or "0 bubble" result as no visible leakage for the minimum test duration, with gas counts translated to "less than 1 bubble per minute" during the test window [S1][S2]. For liquid testing, 1 mL is conventionally equal to 16 drops, which is how small rates are converted for resilient seats [S1]. The practical detection method for small leaks on a gas test is a 2 to 4 mm ID tube connected to the seat cavity and submerged 3 to 6 mm into water, so a human can count bubbles rising from the open end [S1]. Butterfly valves, together with gate, globe, plug, check, floating ball, and trunnion-mounted ball designs, are explicitly inside the API 598 scope [S2][S6].
Soft-seated butterfly valves, those with PTFE, EPDM, or other elastomeric liners, sit in the same "zero leakage" bucket as soft-seated ball valves under API 598, with both the liquid and the air test required to pass clean [S2][S4][S5]. "True zero" is the spec target, not a physical guarantee: small-molecule gases (hydrogen, helium) will still slip past packing and seat edges, which is why the wording is "bubble-tight" rather than absolute zero [S5]. This is also why bubble-tight butterfly seats are usually specified for gas service, pharma, and food-grade systems, while metal-seated butterflies remain common on high-temperature lines where elastomers cannot survive [S3][S4].
Test pressure, duration, and how the procedure runs
API 598 requires two closure checks per valve: a low-pressure seat test in air or nitrogen at about 80 psig (≈5.5 bar) and a high-pressure seat test in water at 1.1× the valve's pressure rating [S1][S4][S8]. The shell test, run before the seat test, holds the body and bonnet at 1.5× rated pressure (water) with no visible leakage permitted through the body, body liner, or body-to-bonnet joint, and no structural damage [S2][S7].
For the seat step, the valve is mounted between test flanges with correctly seated gaskets, fully closed, and pressurised from the upstream side while the downstream side is monitored for leakage [S7]. The test duration is typically 60 to 120 seconds, ambient temperature, and the leakage count is taken as drops per minute for liquid or bubbles per minute for gas, against the size-based limits in the table [S3][S4][S8]. For a reduced-bore butterfly (e.g. a 12×10 NPS wafer lug), the leak rate is keyed to the larger end size (NPS 12), not the bore, unless the purchaser agrees otherwise [S1].
Butterfly-specific failure modes that show up in the seat test
Most butterfly seat-test failures trace back to a short list of mechanical causes. Disc interference with the liner, scored EPDM or PTFE seats from upstream debris, and shaft seal bypassing can each produce leaks that look identical on a bubble count but have very different root causes [S4][S7]. On metal-seated designs, leakage on the seat test is dominated by lapping quality, seat concentricity, and thermal distortion of the disc edge against the seat, with the allowable rate deliberately de-rated for larger NPS because the seating circumference is larger [S2][S4].
Resilient-seated failures are usually one of three patterns: liner swelling or chemical attack (incompatible elastomer in the service fluid), cold flow creep of the PTFE under sustained temperature, or a shifted disc that no longer centres on the seat [S4][S7]. Specifying a higher closure-class butterfly for gas service only pays off if the body and disc geometry actually support the seat, which is why the seat test is run on the assembled valve, not on the seat as a separate part [S2][S7]. For actuator sizing on these bubble-tight designs, expect higher breakaway torque than a comparable Class IV metal-seated butterfly; the elastomer is doing the sealing work and that friction is paid at every cycle, which links directly to the All-Electric vs Pneumatic Actuator: Remote Valve Site Spec selection logic.
Comparison: API 598 vs FCI 70-2 vs ISO 5208 vs API 6D
For a soft-seated butterfly, the practical answer is "API 598 bubble-tight," but it is worth seeing how that maps to the other leakage regimes an engineer will meet on a datasheet: [S4]
API 598 (resilient-seated butterfly): 0 drops/min liquid, 0 bubbles/min gas, all NPS, test at 1.1× rated pressure in water and ≈5.5 bar in air [S2][S4]. API 598 (metal-seated butterfly): size-scaled, 0 to 28 drops/min liquid and 0 to 56 bubbles/min gas [S1][S2]. ANSI/FCI 70-2 Class VI: roughly 0.15 to 0.3 mL/min of air or water depending on port size, used on soft-seated control valves and explicitly the tightest FCI band [S3]. ISO 5208: grades A (tightest) through G (loosest), with Rate A broadly aligned to bubble-tight soft seats [S4][S5]. API 6D: zero visible leakage for soft-seated pipeline valves during the seating test, with full-port metal-seated check valves as a documented exception [S5].
The take-away is that "Class VI" and "API 598 bubble-tight" are not the same test, even though both are colloquially called "zero leakage," because FCI 70-2 is written around control-valve port geometry while API 598 is written around valve type and NPS [S3][S4][S5].
Spec boundaries and what API 598 does not cover
API 598 is a test standard, not a design or face-to-face standard. It is invoked by API 600 (gate), API 602 (compact gate), API 603 (class 150 gate), API 608 (ball), API 6D (pipeline valves), and MSS-SP-61 (general steel-valve pressure testing), and it is the default acceptance test for soft-seated butterfly valves in refinery and downstream work [S2][S4][S5]. It does not define a butterfly's pressure-temperature rating, body wall thickness, or face-to-face dimensions, those sit in ASME B16.34, API 609, and MSS-SP-67 / API 609 category A or B, so a datasheet that quotes "API 609 / API 598" is splitting design and test cleanly [S4][S7].
EN 12266-1 is the European counterpart and overlaps closely with API 598 on closure leakage rates; on a multi-region project it is normal to see both called out, with the tighter of the two governing [S7]. Butterfly valves for cryogenic service, fire-safe service, or fugitive-emission-controlled service (ISO 15848) will stack additional tests on top of API 598, and a seat-test pass on its own does not qualify a valve for any of those duties. The same is true for sour service: NACE MR0175 covers material limits, while API 598 only confirms the assembled valve seals at pressure.
For the process side, a bubble-tight butterfly under API 598 typically pairs with a butterfly valve body rated to ASME B16.34 and is tested per the pumps valves piping envelope. Where the seat test is part of an in-line diagnostic rather than a FAT, a portable measurement test kit with a 2 to 4 mm bubble tube is the field version of the bench procedure [S1].
Trackable signals to watch
Two near-term items will tell you whether the API 598 seat-test language is being applied more strictly on soft-seated butterfly orders: (1) the share of RFQs that explicitly call out "0 drops, 0 bubbles per API 598" in the seat-test line item rather than relying on a generic "test per API 598" note, and (2) whether EN 12266-1 Rate A acceptance is being listed alongside the API 598 line, which is a common indicator of European downstream or EPC work. Both are observable on tender documents, factory acceptance test reports, and the electronic test records that ship with the valve data pack. [S2]