Clean agent suppression and the room integrity door fan test

Clean agent suppression, and why the room is the system

A clean agent fire suppression system floods a protected room with a gaseous agent that puts a fire out, and the room integrity test, the door fan test, is what proves the room will hold that gas long enough to do the job. Water stays where it lands. Gas does not. A clean agent reaches its extinguishing concentration in seconds, then it starts leaking back out through every gap in the enclosure, and the moment the concentration drops below what the fire needs to stay dead, the fire can come back. The protection is the concentration, and the concentration only lasts as long as the room holds it.

That is why the door fan test is the make-or-break step, not a formality at the end. A system can have correct detection, a properly sized cylinder bank, and clean releasing logic, and still fail to extinguish because the room leaks the agent out a cable trench in three minutes. The cylinders fire, the gas floods, and it is gone before the fire is out. A clean agent with no integrity test, or in a room nobody sealed, is a very expensive bang.

The whole guide circles one fact: the agent and the enclosure are one system. Specify the gas all you want. If the room cannot hold it for the required minutes, you do not have suppression. You have a discharge.

How does a clean agent put out a fire?

A clean agent extinguishes by one of two mechanisms, and which one it uses tells you almost everything else about the system. Chemical agents, the halocarbons, put the fire out mainly by absorbing heat. They flood the space and soak up the energy at the flame faster than the reaction can produce it, and the fire goes out at a concentration that still leaves plenty of oxygen in the room for a person to breathe. FK-5-1-12 and HFC-227ea both work this way, with a small secondary chemical effect as the agent breaks down in the flame.

Inert gases work the opposite way. They put the fire out by lowering the oxygen, not by cooling. IG-541, IG-55, and IG-100 are blends of nitrogen, argon, and sometimes carbon dioxide that dilute the room's oxygen from about 21 percent down to roughly 12 to 15 percent, low enough that flame cannot sustain itself but high enough that a person can still function long enough to leave. There is no residue and nothing to clean up with either family, which is the point of a clean agent in a room full of electronics.

The mechanism drives the design concentration, the human-safety limits, the discharge behavior, and the cylinder count, so it is the first decision, not a detail. Heat-absorbing agents reach concentration at single-digit percentages. Oxygen-lowering agents have to fill a third or more of the room volume, which is why an inert gas job carries far more cylinders.

Clean agent types and the HFC phase-down

Clean agents split into the two families above, and inside each family the choice is now as much regulatory as technical. The workhorse halocarbon for years was HFC-227ea, sold as FM-200, a hydrofluorocarbon that works well and stores compactly. The problem is its global warming potential, commonly cited around 3,200, and HFC production is being cut under the AIM Act, which steps domestic production down by roughly 85 percent by 2036. An installed FM-200 system is not illegal, but the agent gets scarcer and more expensive to recharge, which matters when your protection model assumes you can refill after a discharge.

The low-GWP halocarbon is FK-5-1-12, the fluoroketone formerly sold as Novec 1230, with a global warming potential near 1 and an atmospheric lifetime measured in days. It is the common path for a new chemical system. The inert gases sidestep the phase-down entirely, with zero ozone impact and a global warming potential of effectively zero: IG-541 (nitrogen, argon, and carbon dioxide, sold as Inergen), IG-55 (argon and nitrogen), and IG-100 (straight nitrogen).

The cost of inert gas is space and discharge dynamics. It stores as high-pressure cylinders, commonly 200 to 300 bar, and takes far more cylinder room than a halocarbon, and the discharge is loud enough to disturb spinning hard drives, so listed acoustic nozzles matter where rotating media is present. Pick the agent for the room, the recharge model, and the regulatory horizon, not for what the last job used. The broader fire scheme this room sits inside is covered in the data center fire and life safety overview.

What is the design concentration, and is the agent safe to breathe?

The design concentration is the percentage of agent by volume the system is built to reach and hold, and it is the extinguishing concentration plus a safety factor. You start with the minimum extinguishing concentration, the lab-measured percentage that just puts the fire out, then multiply by a safety factor to get the design concentration the room is actually flooded to. Common practice under NFPA 2001 uses roughly a 1.2 factor for Class A surface fires and about 1.3 for Class B flammable-liquid hazards, so the room sees more agent than the bare minimum, with margin for mixing and real-world loss. Verify the exact concentration and factor against the agent listing and the project documents.

The safety question runs the other direction. The design concentration has to be high enough to suppress but low enough not to harm the people who might be in the room. That is the NOAEL and the LOAEL, the no-observed and lowest-observed-adverse-effect levels. For a normally occupied space the design concentration is generally kept at or below the NOAEL, so the agent is safe for the expected exposure. Commonly cited figures put the FK-5-1-12 NOAEL near 10 percent against a design concentration in the 4 to 6 percent range, and the HFC-227ea NOAEL near 9 percent against a design near 7 percent, which is the comfortable margin those agents are known for. Inert gases work by oxygen reduction, so their human limit is the low-oxygen exposure, not a chemical NOAEL.

Treat these numbers as the listing's, not yours. The exact extinguishing concentration, the safety factor, the NOAEL, and the maximum allowable exposure time come from the agent manufacturer's listing and NFPA 2001, and they vary by agent and by edition. Confirm them before you sign off the concentration.

How long must a clean agent hold the concentration?

A clean agent has to hold its concentration at the protected height for a minimum time after discharge, commonly 10 minutes, so the hot surfaces cool below reignition and the fire cannot reflash once the agent starts leaking away. Putting the flame out is the easy part. Keeping it out is the hold. A power supply or a cable that started the fire stays hot for minutes, and if the agent drains before those surfaces cool, the fire comes back in a room that no longer has any suppression left to give.

NFPA 2001 frames this as a retention requirement, not just a discharge requirement. The common target is to hold at least 85 percent of the adjusted minimum design concentration at the highest level of combustibles for a minimum of 10 minutes. Read that carefully, because three things in it trip people up. It is 85 percent of the concentration, not the full design number, because some decay is allowed. It is measured at the highest level of combustibles, the top of the tallest rack or equipment you are protecting, because that is the hardest place to keep concentration as the agent settles. And the 10 minutes is a common minimum, but the AHJ, the hazard, and the time it takes to get a responder to the room can push it longer.

The exact percentage and minutes are edition-dependent and risk-dependent, so confirm the required hold against the adopted NFPA 2001 edition and the project's fire protection basis of design. The number is not really the point. A discharge that floods the room and then loses the gas in three minutes did not protect anything, and the only way to know which you have is to test the room.

What is a door fan (enclosure integrity) test?

A door fan test, also called a room integrity test or enclosure integrity test, measures how leaky a protected room is and from that predicts how long it will hold the clean agent, without ever discharging a cylinder. A calibrated blower is sealed into the doorway. The fan pressurizes the room, then depressurizes it, while gauges read the airflow and the pressure difference across the envelope. From those readings the software calculates the equivalent leakage area of the room, the total of every gap added up as one equivalent hole, then models how fast the agent would drain through it.

The output is a predicted retention time. The model tracks the descending interface, the boundary between the agent-rich layer and the air leaking back in, and reports how long it takes that interface to fall to the minimum protected height. If the predicted hold meets or beats the required hold, commonly the 10 minutes at the protected height, the room passes. If it falls short, the room fails until the leaks are found and sealed and the test is rerun. There is no partial credit. The room either holds long enough or it does not.

This test replaced the old discharge acceptance test, and for good reason. A full discharge proved retention by dumping the entire cylinder bank, which cost a recharge every time, told you nothing until the agent was already gone, and could not be repeated cheaply year after year. The door fan gives you the same answer, in numbers, repeatably, for the cost of a few hours with a blower and a laptop. It is the most useful tool on a clean agent job, and the one most often skipped on a rushed turnover.

Descending interface vs continual mixing

The integrity test reports retention under one of two physical models, and which one applies depends on whether the room keeps mixing after discharge. In the descending interface model, the agent and air separate into layers, a heavier agent-rich layer that slowly sinks as agent leaks low and air leaks high, with a fairly sharp boundary descending through the room. Retention is the time for that boundary to fall to the top of the protected equipment. This is the common model for halocarbon agents, which are heavier than air.

The continual mixing model assumes the room stays stirred, by residual airflow or by an agent and air mix that does not stratify, so the concentration falls fairly uniformly everywhere rather than draining as a descending layer. Inert gas rooms and rooms with residual air movement often sit closer to this model. The reported hold time differs between the two, sometimes substantially, so the test report has to state which model it used and why.

This matters in the field because it is where a report can quietly mislead. A room reported under the more forgiving model can show a passing hold time that the real room, behaving like the other model, will not deliver. The fix is to confirm the technician applied the model that matches the agent and the room, and that the lower-floor and ceiling leakage was measured, not assumed. When a report passes by a thin margin, the model assumption is the first thing to check.

Sealing the room: where the agent leaks out

Every clean agent room leaks where something passes through the envelope, and the integrity test exists to find those paths before the fire does. The usual offenders are predictable. Cable and busway penetrations through the walls and floor, where trays and conduit punch the barrier by the dozen. Door gaps and undercut thresholds. Dampers in the ductwork that do not seat tight. Conduit that runs from the protected room into a panel elsewhere and carries agent out through the raceway. And the two that catch people most: the raised access floor and the suspended ceiling.

The raised floor and the ceiling plenum are not separate rooms. The protected volume includes the space under the access floor and, depending on the design, the space above the ceiling, because agent floods and drains through both. If the underfloor is open to an adjacent room through an unsealed trench, or the ceiling plenum is shared with the corridor, the agent has a highway out and the headline room volume on the drawing is a fiction. Seal the boundary at the actual envelope, top and bottom, not just the walls you can see.

The seal is the same work as firestopping the rated barriers, and it is the same fight. Each penetration gets a proper sealant, collar, or pillow sized to the opening and the penetrant, and the door gets sweeps and seals that actually contact. The part that gets lost: this seal degrades after turnover. Every cable add reopens a path, and the crew that pulled the cable rarely reseals it correctly. That is why the test is repeated, and why a room that passed at commissioning fails two years later with no other change.

Discharge time and how the agent behaves on release

A clean agent does not trickle in. It discharges fast, and how fast depends on the family. Halocarbon agents are stored as a liquid under pressure and are required to discharge quickly, commonly reaching the design concentration in about 10 seconds, because the agent has to get to extinguishing concentration before a fast fire grows past it. That speed is also what spikes the room pressure, which the venting section covers. Inert gases are stored as a high-pressure gas and discharge more slowly, commonly over roughly 60 to 120 seconds, because pushing that volume in faster would overpressure the room.

The fast halocarbon discharge has a side effect worth knowing. As the liquid flashes to vapor it absorbs heat and the room can fog and chill briefly, and the rapid vaporization pulls the room pressure negative for an instant before the vapor flow pushes it positive. Inert gas, going in as gas the whole time, mostly just pushes the pressure up. Either way the discharge is loud, and for inert gas the nozzle noise alone has been shown to disturb and even damage spinning hard drives, so listed quiet nozzles and placement matter where rotating media is present.

The discharge time also sets the clock on everything that has to happen first. Fans have to be off and dampers closed before or as the agent releases, the room has to be clear of people, and the abort window has to have expired. The agent reaching concentration is the end of a sequence, not the start of one.

Why does my clean agent room need pressure venting?

A clean agent room needs pressure relief venting because the discharge itself can spike the room pressure hard enough to damage the walls, the door, or the structure, and a tighter room makes it worse, not better. This is the tension at the heart of clean agent design. You want the room tight so it holds the agent for the retention time. But the tighter the room, the less the discharge pressure has anywhere to go, so the peak pressure climbs as leakage falls. Seal the room perfectly and you can blow the drywall off the studs or burst the door on discharge.

The fix is a pressure relief vent, a damper or louver that stays shut in normal operation and opens only during the pressure spike of a discharge, then reseats. It gives the discharge somewhere to dump the brief surge without leaving a permanent hole that would drain the agent. That lets the room be sealed tight enough for the hold time and still survive the discharge. The vent has to be sized to the agent, the discharge rate, and the room, by calculation, and halocarbons need both the negative and the positive pressure swing assessed because the discharge pulses both ways.

The mistake is to assume the room leakage doubles as pressure relief. Sometimes a leaky room does relieve enough that no vent is needed, but a properly sealed room usually does not, and the same sealing that earns the hold time is what creates the pressure problem. Run the peak pressure analysis. NFPA 2001 requires the discharge not to damage the enclosure, and a room sealed for retention with no relief path is exactly the room that gets hurt.

HVAC shutdown and damper closure

The cooling system and the suppression system are wired together because air movement decides whether the agent stays in the room. On a confirmed alarm the air handlers serving the protected space shut down and the dampers close, before or as the agent discharges. If the fans keep running, they sweep the agent out of the room and pull fresh air in, and the concentration you spent a cylinder bank to reach is gone in seconds. A clean agent discharged into a room with the cooling still running is a discharge into the return duct.

Dampers do the containment. Smoke and isolation dampers in the supply and return close on the fire signal to seal the room to the volume the agent calculation assumed, and a damper wired to close but rusted open is a leak path the integrity test will find and the discharge will use. The damper closure shows up in two documents that have to agree: the fire cause-and-effect matrix and the mechanical sequence of operations. If they disagree, one of them is wrong and the room is not protected.

Because the two systems are coupled this tightly, they get commissioned against each other, not in isolation. You cannot prove the suppression without driving the cooling system through its fire response and watching the fans actually stop and the dampers actually seat. The airflow and cooling side is covered in the data center cooling and airflow overview, and the detection and cause-and-effect side in the data center fire and life safety overview. The point here is narrow: the agent does not hold in a room whose fans never stopped.

Detection, the time delay, and the abort

A clean agent is released by the detection and releasing system, and the logic is built so a single faulty detector never dumps the room. Detection is cross-zoned, meaning two independent detection zones have to agree before the system arms to discharge. A first detector going into alarm raises the warning and tells someone to investigate. A second, independent detector confirms it, and only then does the releasing sequence start. That two-detector logic is what keeps a dusty smoke head or a single fault from discharging a six-figure cylinder bank into an occupied room.

Once both zones confirm, a pre-discharge time delay runs before the agent releases, commonly on the order of 30 seconds, with a horn and strobe warning the room. The delay does two things. It gives people time to leave, and it gives an operator who knows it is a false alarm time to hit the abort. The abort station holds the discharge while it is pressed or for as long as the design allows, and the better designs do not let abort cancel the event outright, only delay it while someone deals with the cause. There is also a manual release, a pull station that fires the system immediately for someone who sees the fire and does not want to wait for the detectors.

This logic lives in NFPA 72 and in the cause-and-effect matrix, and it is only real if the integrated test proves it. Trip both zones, watch the delay run, test the abort, test the manual release, and confirm the releasing panel actually energizes the cylinder solenoid at the end of the count. The full detection and alarm scheme is covered in the data center fire and life safety overview.

Agent storage: cylinders, weight, and the hydrostatic clock

The agent lives in cylinders, and the one thing that has to be true is that the agent is actually still in them. A slow leak past a valve empties a cylinder over months with no alarm, and a system that discharges a half-empty bank never reaches design concentration. So the cylinders get checked on a schedule, and the check is physical, not a glance at a panel light.

Halocarbon cylinders are weighed, because the agent is a liquid and weight tells you how much is left. The common rule is to weigh on a periodic cycle, often every 6 months, and if a cylinder has lost more than about 5 percent of its agent or 10 percent of its pressure, it goes back to the shop to be refilled. Inert gas cylinders hold a gas, not a liquid, so weight does not tell you much. They are judged by pressure read off the cylinder gauge, with the same out-of-service trigger if the pressure has fallen. Liquid weight checks and pressure checks confirm the same thing in different ways: that the design quantity is present.

The cylinders themselves are pressure vessels under federal rules, so they carry a hydrostatic retest clock independent of the fire code. Under U.S. DOT rules a cylinder is commonly retested on a 5-year interval once it has been discharged or emptied, with a longer interval, often cited at 12 years, for a cylinder that has stayed sealed and never discharged. Confirm the interval against the current DOT regulation and the cylinder's own markings. A cylinder past its hydrostatic date cannot be legally refilled, which becomes a problem the day you need to recharge after a discharge.

Clean agent vs inert gas: which for the room?

They are different tools, and the room usually decides, not preference. Halocarbon agents like FK-5-1-12 reach concentration at single-digit percentages, so they need far less storage, discharge in about 10 seconds, and fit a tight cylinder room. That makes them the easy choice where space is short and the recharge model can absorb the agent cost. The catch is the agent cost and, for the older HFCs, the phase-down that makes recharge scarcer over time.

Inert gases reach concentration by displacing oxygen, so they have to fill a third or more of the room volume and carry many more cylinders, often a whole separate room of them, with a slower 60 to 120 second discharge and a louder release. What you buy for that is an agent with effectively zero global warming potential, no decomposition products, and no phase-down hanging over it, stored as plain nitrogen and argon. On a large protected volume the inert cylinder count can be the deciding constraint, and on a room full of spinning disks the discharge noise can be too.

The honest answer is that both extinguish, both leave no residue, and the choice comes down to available cylinder space, the recharge and regulatory horizon, discharge noise near rotating media, and the project budget. What does not change between them is the enclosure. Either agent leaks out of a room that was not sealed and tested, so the door fan test and the seal package matter exactly as much regardless of which gas the brochure favors.

Commissioning and acceptance

A clean agent system is accepted by proving each piece and then proving the sequence, and almost none of it involves discharging agent. The acceptance set runs in a deliberate order. The room integrity door fan test confirms the enclosure holds the agent for the required time. The releasing panel functional test confirms the panel sees both detection zones, runs the time delay, honors the abort and the manual release, and energizes the cylinder release circuit at the end of the count, usually with the cylinders fitted with a test indicator rather than live actuators for the dry run. The cause-and-effect test drives every input and watches every output: fans stop, dampers close, warning devices sound, notification reaches the panel and the monitoring station.

The piping gets a puff test, not a discharge. A small charge of nitrogen or another test gas is pushed through the distribution pipe and out each nozzle to confirm the pipe is clear, continuous, and not blocked, and that every nozzle is open and pointed where the drawing says. It proves the agent will actually get to the room and out the nozzles when the day comes, without spending a cylinder to find a capped tee.

A full discharge is generally not part of acceptance. It is expensive, it puts the agent and the recharge clock to work for nothing, and the door fan test plus the functional and puff tests prove the same things without it. A discharge test is reserved for the rare case the AHJ or the design specifically calls for one. If someone proposes a live discharge as routine commissioning, ask why the door fan test will not answer the question instead.

What periodic testing does a clean agent room need?

A clean agent room is not done at turnover, because the enclosure changes the moment people start working in it. The recurring program runs on a few clocks. The room integrity door fan test is commonly repeated annually, because every cable pull, every new penetration, and every propped ceiling tile reopens a leak path, and a room that held 11 minutes at commissioning can fall under 10 after a year of moves and adds. The retest is the only thing that catches that drift, and it is the test most often dropped from the maintenance contract to save a line item.

The cylinders get the storage checks on their own cycle, commonly a 6-month weight or pressure check and the DOT hydrostatic retest at the longer interval. The detection and releasing get functional testing under NFPA 72, including the cross-zone logic, the time delay, the abort, and the manual release, commonly annually. The cause-and-effect gets re-proven on the same cadence, because a control change or a damper failure since the last test breaks the sequence quietly.

The intervals come from NFPA 2001, NFPA 72, the manufacturer's listing, and the DOT cylinder rules, and they shift between editions, so build the calendar from the adopted editions and the equipment manuals, not from memory. The owner who skips the annual door fan test still has a cylinder bank and a panel that test fine, and a room that no longer holds the agent. The hardware passing tells you nothing about whether the room still does.

What to document

The clean agent record is what proves the room was protected and what the next technician maintains against, so it has to capture the enclosure, not just the hardware. The integrity test is the centerpiece, because it is the one number that ties the agent to the room. Record the room and its protected volume, the agent and design concentration, the required hold time, the door fan predicted hold and which interface model it used, the pressure relief provision, the cylinder status, and the pass or fail with the date.

Common mistakes

• Specifying a clean agent and never running the door fan test, so nobody knows whether the room holds the agent.
• Leaving new cable penetrations unsealed after a move or add, draining the agent through a gap that was not there at commissioning.
• Sealing the room tight for retention and providing no pressure relief vent, so the discharge damages the walls or the door.
• Counting on the room's leakage as pressure relief without running the peak pressure analysis.
• Wiring the fan shutdown and dampers to the alarm and never proving they actually stop and seat on discharge.
• Excluding the raised floor or ceiling plenum from the sealed envelope, so the real protected volume leaks faster than the drawing assumes.
• Never weighing the halocarbon cylinders or pressure-checking the inert cylinders, so a slow leak empties the bank unnoticed.
• Treating the design concentration as fixed without checking it stays at or below the NOAEL for an occupied room.
• Accepting a door fan report that passed under the more forgiving interface model without confirming it matches the agent and room.
• Skipping the annual integrity retest, so the room drifts leaky between turnover and the day it is needed.

Field checklist

Standards and references

Clean agent systems are governed first by NFPA 2001, the Standard on Clean Agent Fire Extinguishing Systems, which sets the design concentration and safety factors, the agent quantity, the retention and the room integrity requirement, the pressure relief and discharge criteria, and the inspection and testing cadence after turnover. The agent listing, from the manufacturer and the listing laboratory, fixes the minimum extinguishing concentration, the NOAEL and LOAEL, and the maximum exposure for that specific agent. Treat the listing as governing the agent-specific numbers and NFPA 2001 as governing the system.

Detection and releasing follow NFPA 72, the National Fire Alarm and Signaling Code, which governs the cross-zoned detection, the releasing service, the time delay, the abort, and the cause-and-effect sequence that ties the discharge to the fans and dampers. The cylinders are pressure vessels under the U.S. DOT regulations in 49 CFR, which set the hydrostatic retest intervals independent of the fire code. The HVAC interaction draws on the mechanical design and, by topic, ASHRAE guidance for the airflow and the equipment the fire response has to shut down, while the upkeep of the releasing and control gear tracks the manufacturer's instructions and NFPA 70B-style electrical maintenance practice.

Edition numbers and clause references change every cycle, the jurisdiction adopts a specific edition that is often a few cycles back, and it amends locally, so confirm the adopted editions and any local amendments against the project documents before citing a clause, and let the AHJ settle any conflict. This room is one part of a larger fire scheme covered in the data center fire and life safety overview.

Units, terms, and acronyms

Clean agent work carries its own vocabulary, and the same idea reads differently across an NFPA 2001 calculation, an agent cut sheet, and a door fan test report. The terms below are the ones that travel across the whole job.

Clean agent

A gaseous fire-extinguishing agent that leaves no residue, either a heat-absorbing halocarbon or an oxygen-lowering inert gas

Design concentration

The agent percentage by volume the room is flooded to, the minimum extinguishing concentration times a safety factor

MEC / MDC

Minimum extinguishing concentration, the percentage that just puts the fire out, and minimum design concentration after the safety factor

NOAEL / LOAEL

No-observed and lowest-observed-adverse-effect levels, the human-exposure limits the design concentration is held against

Retention / hold time

The minimum time the room must hold the agent at the protected height, commonly at least 10 minutes under NFPA 2001

Door fan / room integrity test

A pressurization test that measures enclosure leakage and predicts how long the room will hold the agent

Equivalent leakage area (ELA)

The total of all the room's gaps expressed as one equivalent opening, the door fan test's core measurement

Descending interface

The boundary between the agent-rich layer and infiltrating air that sinks as the room leaks, used to model retention

Pressure relief vent (PRV)

A vent that opens only during the discharge pressure spike so a sealed room is not damaged

IG-541 / FK-5-1-12

IG-541 is an inert nitrogen, argon, and carbon dioxide blend; FK-5-1-12 is a low-GWP fluoroketone halocarbon

Cross-zoned detection

Two independent detection zones that must both confirm before the system releases agent

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What fire suppression do data centers use?
• What is bypass airflow in a data center?
• What fire suppression do data centers use?
• Why do battery rooms need ventilation?
• What pressure do you test chilled-water piping to?